Limetech un-Raid Media Storage Server support thread [Text View]

limetech

08-29-05, 10:28 PM

These are my guesses from working with other raid systems...

Well, since each drive is its own volume, I would guess the performance is the same or worse than just a standalone disk. The disks are not spanned so there is no performance gain in reads by using this un-RAID. Writes would be slower than a single disk since parity would have to be calculated. I don't know if they mentioned how they perform parity (hardware vs. software), but since theyare using Linux, I'm guessing software raid.

Performance COULD be better if you were able to predict usage patterns and store files that would be accessed simultaneously on different disks. A modern single disk could handle multiples DVD requests, but it all depends on the demands of disk and performance. I didn't see any benchmarks on their website.

There's potential for energy savings and disk longevity by spinning down disks, but they using an intel Celeron proc still uses too much wattage for my liking (for something that is on 24/7)

Right on all counts :)

Some details... First, remember this system is designed for media data storage where ease of use, expansion, and low chance of complete data loss are very desireable features. Since it's a network attached device, the bottleneck is almost always going to be the network and/or the device from which the media data is being read from.

Regarding energy savings... We spin the disks down as much to save wear and tear on the disks as save energy (and reduce heat). We're working on a feature to cause the system to suspend after a period of inactivity, and then either wake up on it's own at a certain scheduled time or wake up via "wake-on-lan".

albrigsr

08-30-05, 12:38 AM

Tom,

Thank you for the reply. I sent an email the other day. Please let me know if you didn't get it (I sent to the Lime Technology address) and I can send another one.

Sincerely,
Scott Albright

madpoet

08-30-05, 06:50 AM

It's impressive, but honestly pretty expensive. Considering the ability to (freely) use software RAID5, it's a hard sell.

limetech

08-30-05, 04:32 PM

It's impressive, but honestly pretty expensive. Considering the ability to (freely) use software RAID5, it's a hard sell.

Yes software RAID is free, but you still need a system to run it on, and there are some significant disadvantages to RAID-5 for media storage applications.

Using retail prices for hard drives, by the time you've filled up the system, you are well under $1/GB. If you buy drives over time making use of sales and rebates, you can approach 50 cents/GB - ok, maybe 75 cents :) This is extemely competitive with other RAID NAS devices on the market.

I realize most people on this forum "build their own" (as do I). My motivation for starting this thread is not to drum up sales necessarily, but to have our approach critiqued by people I know are knowledgeable.

stanger89

08-30-05, 04:38 PM

FrothyDog

08-30-05, 04:44 PM

limetech

08-30-05, 04:53 PM

I think what madpoet is getting at is the cost (without drives) is the price is rather high for your software, considering the cost of the hardware. I was pretty interested until I saw the pricetag. (Coming from a guy who has an 8 drive + 3ware RAID-5 array).

Fair enough. Actually, a big part of the cost is assembly and a shipping container. We don't have the volume (yet ;) ) to have them mass assembled.

Out of curiosity, if you were to build your current array using a 3ware card, mobile racks for the drives, and a high quality case and power supply, what would it cost you per GB of storage (and no fair using an old motherboard you have laying around)?

limetech

08-30-05, 04:56 PM

I had to laugh when I saw the picture.... It looks exactly like my media server. I have the same Coolermaster Stacker case with 8 of the Cremax hot-swap drive bays installed. It looks pretty slick.... of course you already knew that.

Eric

Isn't this the best case you have ever owned? It's so easy to work on and they include a lot of nice touches like the wheels and connector for hooking up two power supplies.

stanger89

08-30-05, 05:41 PM

Fair enough. Actually, a big part of the cost is assembly and a shipping container. We don't have the volume (yet ;) ) to have them mass assembled.

Granted.

Out of curiosity, if you were to build your current array using a 3ware card, mobile racks for the drives, and a high quality case and power supply, what would it cost you per GB of storage (and no fair using an old motherboard you have laying around)?

I'll answer this without discussing drive prices because you sell your system without drives, so we'll compare apples to apples.

I've got a 3ware 7506-8 RAID card, I can build (as near as I can tell without knowing your exact part numbers) your hardware from newegg + that RAID card for $1060.42. Now remember than includes a RAID card that's $395 (the 12-port version is about $200 more so a whole 3ware 12-port setup would be about $1300). So that means I can buy your hardware for $665.42, basically half what you're charging. That would be the PATA config, for the SATA config (3ware 9500s) in the 12 port, that would be just over $1350.

Now take it as constructive criticism, because again, it is an interesting setup, especially with the parallels to the Kaleidescape system (which is also a RAID-4 or similar setup). I think if you just sold the flash drive with your software (and an appropriate list of supported configurations) for something in the neighborhood of $50-100 you'd get a lot of bites from those like us.

-edit

FWIW, I already had a donor PC, and built my array when 250GB drives were the sweetspot price wise (about 9 months ago), my array (3ware+drives) cost me ~$1300, but 250GB drives were running 125-150 at the time.

-edit 2

Also my array is nearly full so I'm keeping an eye out for attractive storage solutions. :)

limetech

08-30-05, 06:31 PM

...So that means I can buy your hardware for $665.42, basically half what you're charging...

Hmm, our h/w costs more than that & we use volume pricing...

Now take it as constructive criticism, because again, it is an interesting setup, especially with the parallels to the Kaleidescape system (which is also a RAID-4 or similar setup). I think if you just sold the flash drive with your software (and an appropriate list of supported configurations) for something in the neighborhood of $50-100 you'd get a lot of bites from those like us.

Thank you. This is the kind of feedback I'm looking for. We can definately offer the software on a Flash drive, and not only that, provide detailed assembly instructions on our website for those who what to "roll their own".

-edit

FWIW, I already had a donor PC, and built my array when 250GB drives were the sweetspot price wise (about 9 months ago), my array (3ware+drives) cost me ~$1300, but 250GB drives were running 125-150 at the time.

-edit 2

Also my array is nearly full so I'm keeping an eye out for attractive storage solutions. :)

The 320GB PATA drives are currently the "sweet spot". One think nice about our system is that you can upgrade drives in place. So in a few months you could replace those 250's with 500's.

madpoet

08-30-05, 06:47 PM

stanger89

08-30-05, 07:27 PM

Hmm, our h/w costs more than that & we use volume pricing...

Kind of like local PC shops, they usually have a really hard time competing with online (newegg) as well.

Thank you. This is the kind of feedback I'm looking for. We can definately offer the software on a Flash drive, and not only that, provide detailed assembly instructions on our website for those who what to "roll their own".

There will always be a market for the ready-to-go stuff, but I'd guess that you'd get a lot better (ie more sales) response from AVS-types with a "roll your own" option. :)

The 320GB PATA drives are currently the "sweet spot". One think nice about our system is that you can upgrade drives in place. So in a few months you could replace those 250's with 500's.

I've been back and forth on that, and come to the conclusion that (no offense) that "benefit" isn't one. I mean, sure you can upgrade to larger drives, but then what do you do with the old ones? Sell them for pennies on the dollar? I figure, you'll be upgrading due to increased storage need, so it's more logical to keep what you have, and just add more :)

albrigsr

08-30-05, 08:17 PM

I appreciate reading all of your comments. I tried to price out a somewhat similar system and came up with higher than $665 but less than the offered price of $1300 (I just wanted a rough range). I do tend to take on many of these projects myself and consider myself a fairly savvy DIY'er but don't know much about Linux, Samba, etc. so the ready-made version sounds great to me. I had an earlier RAID system (albeit RAID 0 for speed) which worked for a while but ended up with some problems so I am looking for a turn-key system that has some offered support. It looks like this product will work great. I personally like the advantage of using different sized HDs as I probably have 3 250GB, 3 300 GB and 2 400 GBs. I would lose out moving to a RAID 5 system. I also haven't seen a RAID 5 system (pre-built) that doesn't cost a great deal more or is expandable to 11-12 Hard Drives. I do tend to agree that I probably plan to add rather than get rid of old drives though (or I wouldn't need all this space in the first place!). I plan on buying a couple of additional drives (probably 400 or 500 GB) and putting them in the array and copying from the aforementioned drives and moving them 1 by 1 into the array. Anyway, I think I'll take the plunge and I'll be sure to post my experiences on the board.

BTW, I would think that more and more people will ultimately want to move to some backup solution/redundancy solution in the future as people become more dependent on computers and the personal information they contain (Downloaded Music, Digital Pictures, Digital Video, etc.) Just my thoughts!

The only concern I have is whether my modded XBOX (or any other media client like the DSM-320, etc.) will be able to correctly access the server as it is Linux based but assume that I won't have a problem as SAMBA/SMB will provide any Windows based client the ability to do so.

stanger89

08-30-05, 08:47 PM

Samba is just "Windows File Sharing" for linux.

pischke

08-30-05, 08:56 PM

Thank you. This is the kind of feedback I'm looking for. We can definately offer the software on a Flash drive, and not only that, provide detailed assembly instructions on our website for those who what to "roll their own".

I would definitely be interested in the "software and assembly instructions" package - much easier to sneak past the wife... :)

-pischke

SHS

08-30-05, 10:49 PM

SHS

08-30-05, 10:56 PM

So you sneak that past the wife aaa pischke ok it your head not mine :D then look here http://www.serverelements.com/.

stanger89

08-30-05, 11:14 PM

Well ask me that not bad price when other that includes bare chassis, no hard drives run in $4Grand.
If look on FAQ is say Promise controller I wonder card it be

COOLER MASTER CM Stacker STC-T01 $159
Intel D865GLCLK $110
Intel 2.26GHz Celeron D 315 Prescott 533MHz FSB Socket 478 Processor $69
512MB about $50 for good set memory from Crucial, Corsair or OCZ
12 HotSwap Removable Disk Trays Cremax MB123AK cost $18 each = $216
May guest is has possable has 2 or 3 PCI Promise SuperTrak or custom SCSI/IDE card which be any form $350 to $500.
Look like a Sparkle FSP300-60PN ATX 300W $33 each = $66
My guest is base on Server Elements which could custom ver so that $25 there
And USB Flash which what $20 tops
Stanger89 how fig that to high

They're asking $1300 without HDDs. And it doesn't have an expensive RAID card since they're doing a fancy non-RAID partity thing. My guess would be something like the SATAII150 SX8 or FastTrak TX4000, which are in the $100-200 range. Also note in the FAQ that they imply they use the motherboard controller for 4 drives, hence only need to do 8 with the other.

SHS

08-30-05, 11:38 PM

I would count on that SATAII150 SX8 work on the Intel D865GLCLK look at back of PCI slot but it could very well be the FastTrak TX4000 card but any case let add "12" 500 GB hardrive at $317 each = $3804 and one Un-RAID $130 for $5104 which is still $4500 less then custom rack ver U3/4 model.

The motherboard controller for 4 IDE drives is not RAID Controller Compatible it is just a Stardard IDE and only the SATA Controller is RAID Compatible I know my brother has that very same board.

sangpeiris

08-31-05, 02:14 AM

limetech

08-31-05, 04:39 AM

I would GLADLY buy and try your software on a flash card in that range. I've already done it once with a competitor product for $150 and it worked very nicely, but had more limitations than yours seemed to. I would also urge you to consider it.

Our original business model was to offer just the software on a Flash drive.

But as we started testing various motherboards it became apparent that each one had some "quirk" which would present problems to the end user. For example, whenever you add a new drive to the system, some m/b's "reorder" the boot drive sequence so that the Flash drive isn't tried first. This requires the user to hook up a keyboard and VGA monitor and go into the bios. (But not all m/b's have this problem).

Some m/b's had unreliable drivers for the on-board GigE port.

So we made a decision as a company to offer a pre-built product first. This way we could have a controlled platform and make the system extremely reliable.

Having said this, we are considering publishing a "reference design" which lists the key parts required for reliable operation, and making the software available separately.

limetech

08-31-05, 04:44 AM

...I mean, sure you can upgrade to larger drives, but then what do you do with the old ones? Sell them for pennies on the dollar? I figure, you'll be upgrading due to increased storage need, so it's more logical to keep what you have, and just add more :)

This feature is primarily a benefit if you are ugrading really small drives. For example, one customer bought 4x 400GB seagate drives, and then put in a couple 80GB drives he had in an old system. At some point he may very well want to upgrade those.

limetech

08-31-05, 05:02 AM

... I personally like the advantage of using different sized HDs as I probably have 3 250GB, 3 300 GB and 2 400 GBs. I would lose out moving to a RAID 5 system. I also haven't seen a RAID 5 system (pre-built) that doesn't cost a great deal more or is expandable to 11-12 Hard Drives. I do tend to agree that I probably plan to add rather than get rid of old drives though (or I wouldn't need all this space in the first place!). I plan on buying a couple of additional drives (probably 400 or 500 GB) and putting them in the array and copying from the aforementioned drives and moving them 1 by 1 into the array. Anyway, I think I'll take the plunge and I'll be sure to post my experiences on the board.

RAID-5 systems are great for many applications, but they are expensive and very difficult to expand.

BTW, I would think that more and more people will ultimately want to move to some backup solution/redundancy solution in the future as people become more dependent on computers and the personal information they contain (Downloaded Music, Digital Pictures, Digital Video, etc.) Just my thoughts!

Right you are! We see this as a large almost untapped market (as do many other companies).

The only concern I have is whether my modded XBOX (or any other media client like the DSM-320, etc.) will be able to correctly access the server as it is Linux based but assume that I won't have a problem as SAMBA/SMB will provide any Windows based client the ability to do so.

Well SAMBA was first created (http://us3.samba.org/samba/docs/SambaIntro.html) by reverse engineering Windows network file sharing. These days you occasionally do see some interoperability issues, usually because the samba team does a better job at implementing windows networking than M$ does :)

BTW, you can access our system via Mac's also.

limetech

08-31-05, 05:25 AM

Accordin to one of the FAQs, the controller is a Promise Card...

Q: My system takes longer to boot with four drives than it does with five drives, what's happening?
A: If you only have disks in the top 4 slots, the Promise controller bios takes longer to decide there are no hard drives attached to it.

Not much gets past you guys :D

Let me put the mystery to rest. We use simple (though reliable) Promise Ultra 100 TX2 interface cards (or the Ultra 133 version, whichever we can get better pricing on).

For those of you who care: What we did was write a non-striping RAID-4-like driver for linux, along with associated management software. We're using the 2.4.29 kernel and have managed to shrink the system down to around 60MB (but it's compressed on the flash).

So, no h/w RAID card. But realize this: the "h/w" in "H/W Raid" is actually just firmware and maybe a h/w xor engine. The data still has to stream on and off the disks, and that's always the bottleneck-of-last-resort. What h/w Raid is really useful for is striping.

I'm not trying to bash h/w Raid at all. It definately has it's uses and applications. But for a dedicated network attached file server with only a single GigE port (or even 2 ports), it's simply not required.

musicmann

08-31-05, 04:58 PM

Wow! I had been wondering what I was going to do when I run out of space on my 2.5TB server, and this looks like the answer! I'd be interested in a software-only option as I like to roll my own.

The things I'd like to see are:
- Security
- The ability to define the number of parity drives (i.e., if I want tolerance for 2 drives dying with no data loss, that would be sweet).
- The ability to control more than 12 drives
- The ability to "consolidate" the drives into one appearance on the network (I think an old product called CyberNAS could do that)

When I originally got into the HTPC scene, I was looking at linux for my server o/s, but I was having problems copying my DVD ISOs (> 4 or 5 GB) from my Windows box to my linux box. Are large file transfers a problem with this o/s or file system?

Stereodude

08-31-05, 08:10 PM

RAID-5 systems are great for many applications, but they are expensive and very difficult to expand.
They aren't cheap, but they can be easily expanded with the Broadcom BC4852 RAID card.

albrigsr

08-31-05, 08:41 PM

OK. I just placed my order earlier and am taking the plunge. I will post my experiences here on the site.

madpoet

08-31-05, 09:10 PM

Welcome, Blind Guinea Pig 1! ;)

galileo2000

08-31-05, 11:32 PM

limetech

09-01-05, 03:00 AM

...The things I'd like to see are:
- Security
- The ability to define the number of parity drives (i.e., if I want tolerance for 2 drives dying with no data loss, that would be sweet).
- The ability to control more than 12 drives
- The ability to "consolidate" the drives into one appearance on the network (I think an old product called CyberNAS could do that)

Security is in the next software release.

Multiple parity drive support (sometimes called RAID-6) is not going to happen any time soon :)

Array sizes beyond 12 drives is not easily done.

Regarding consolodation... I'm not exactly sure what you mean here. Currently each disk shows up as a separate share, e.g., for a server named "tower" you will see:
//tower/disk1
//tower/disk2
...
//tower/disk11

We've considered "consolodating" these under one directory in order to de-clutter My Network Places, e.g, using a share named "storage":
//tower/storage

with disks appearing as:
//tower/storage/disk1
//tower/storage/disk2
etc.

Is this what you had in mind?

When I originally got into the HTPC scene, I was looking at linux for my server o/s, but I was having problems copying my DVD ISOs (> 4 or 5 GB) from my Windows box to my linux box. Are large file transfers a problem with this o/s or file system?

We move 8GB files back and forth between Windows XP and the UnRaid server in order to measure perfomance. I know we have a 2TB max disk size (due to version of linux we use). I'm not sure what the max file size is, if any. I'll look into that.

limetech

09-01-05, 03:08 AM

They aren't cheap, but they can be easily expanded with the Broadcom BC4852 RAID card.

Yep, that's a pretty powerful card. I couldn't tell from the documentation, but you probably have to use the same size hard drives to expand with (or larger, but not be able to use the storage until you can create a separate array out of it).

albrigsr

09-01-05, 03:12 AM

limetech

09-01-05, 03:12 AM

OK guys my apologies in advance for being stupid and brutal.

I have Sony CX777ES 400 DVD changer controlled from the HTPC through RS-232 serial port. I have immediate access to any of my DVD's with one "Enter" command from my IR keyboard.

400x4.3GB = 16 TB.

Oh yeah, ffdshow..

Once again, I apologize and please correct me.

"Make everything as simple as possible, but not simpler." -- Albert Einstein.

For many Home Theater applications, a DVD changer is definately the way to go.

Thomas J. Coyle

09-01-05, 09:57 AM

galileo2000

09-01-05, 10:43 AM

Galileo,

I get your point that a HD based media server is not the most cost effective solution (I'm pretty sure that's what you meant). That may be true. I personally will be using this not only as a media server but as a true file server for the computers throughout my house (3 different computers that are always on...a few others that aren't used regularly as well). This solution provides me with redundancy which is something I was needing for digital photos, digital music, digital video (home video), etc. I spent countless hours ripping my large CD collection and don't want to do it again. It's great for other data as well (although I am able to back it up easier due to its smaller size). Plus, I have learned to hate sitting through the DVD FBI Warnings, Menus, etc. It's nice to be able to click on a movie and have it immediately start. That doesn't mean that I plan or have ripped all my DVDs to HD because as you have mentioned it's quite expensive. Also, what's the ability to daisy chain the DVD changer? I guess you could install another card with another serial port to do it, right? I'm just asking as I personally own over 700 DVDs so the 400 DVD changer won't help. I've also read various +/- about changers regarding scratched discs, slow reponse, etc. Not to say that your model has those issues. Another big issue would be the ability to serve the stream to multiple clients which a DVD changer cannot do. Your solution locks you into just one television, right? Anyway, both are certainly viable and much better than the alternative of what I have now!

Albrigsr,

I get your point and respect your needs which might be different than mine.

I also did not want to discard this fine product OP has brought here and I am sure some people do need it and will be happy with it.

Couple of points though:
- you cannot daisy-chain Sony changers. However, Sony has 3 different sets of IR codes, so if you have 3 inputs in your receiver you can have 3 CX777ES which gives you 1200 DVD capacity control according to specs from Sony. Also, in order to control it through RS-232 you want to buy another card which gives you couple more serial ports. I haven't done it because I didn't need it but I might in the future, especially if Sony comes out with the newer version of CX777ES.

- I also spent countless hours ripping my CD's. I then reencoded ALL of them with DVD Architect into AC-3 ( and boy, was it a piece of work!). Now I have about 300 CD's stored on 6 DVD's with visual controls and access to any particular CD from the on-screen picture. And guess what, those 6 DVDs are in Sony and Sony plays them :D

However, there are some drawbacks in my solution as well. CX777ES refuses to play PAL. I have to re-encode stuff from PAL to NTSC. As I mentioned, I couldn't use ffdshow. Not a big deal for me, I am quite happy with the Sony picture, but for some here it can be a major turn-off.

galileo2000

09-01-05, 10:45 AM

"Make everything as simple as possible, but not simpler." -- Albert Einstein.

For many Home Theater applications, a DVD changer is definately the way to go.

limetech,

I do appreciate your open mindness.

I am also confident that you have a fine product and there certainly is a market for it for the AVS enthuziasts.

Troubleshooter

09-01-05, 11:32 AM

dpeart

09-01-05, 01:57 PM

Perfect timing for me as I'm looking into a system, and almost pulled the trigger on a RAID5 system with hardware OCE. My biggest concern was the OCE. If you get an additional drive that is a few bytes smaller than the originals, your out of luck.

A couple questions:

1. How do you handle rebuilding? With different size drives, do you have to replace with the same size drive to rebuild, a drive large enough to hold whatever was on the dead drive, or larger drive?

2. Do you support having a backup drive in the system ready to be swapped in as a replacement. If so how does question 1 affect the choice of this drive?

As far as buying a system spec and software on flash drive for $100-$150 that is about what I would spend on XP Pro if I were to build my own RAID5 system, so it would be a no brainer for me. I would buy your solution.

musicmann

09-01-05, 04:25 PM

Array sizes beyond 12 drives is not easily done.
That's too bad. I'm sure a number of home rollers like really high density...I have 13 HDDs in my CM Stacker (plus 1 optical, 1 card reader, and the power panel).

We've considered "consolodating" these under one directory in order to de-clutter My Network Places, e.g, using a share named "storage":
//tower/storage

with disks appearing as:
//tower/storage/disk1
//tower/storage/disk2
etc.
Exactly.

We move 8GB files back and forth between Windows XP and the UnRaid server in order to measure perfomance. I know we have a 2TB max disk size (due to version of linux we use). I'm not sure what the max file size is, if any. I'll look into that.
That's probably all I need to know. I was pretty sure that my problem was just with configuration and possibly file system (ext3). I can roll my own H/W, but I wasn't going to take that "roll a custom kernel" plunge.

Ken Greene

09-01-05, 06:12 PM

how about a Rack mounted version????

or i can roll my own if you do offer the software only option

albrigsr

09-01-05, 06:22 PM

shokunin

09-01-05, 08:28 PM

OK guys my apologies in advance for being stupid and brutal.

I have Sony CX777ES 400 DVD changer controlled from the HTPC through RS-232 serial port. I have immediate access to any of my DVD's with one "Enter" command from my IR keyboard.

400x4.3GB = 16 TB.

Oh yeah, ffdshow..

Once again, I apologize and please correct me.

I think you missed a decimal place :)... more like 1.6TB. DVD changers are fine as long as they work and you can deal with the wait time between switching disks.

shokunin

09-01-05, 08:33 PM

Our original business model was to offer just the software on a Flash drive.

But as we started testing various motherboards it became apparent that each one had some "quirk" which would present problems to the end user. For example, whenever you add a new drive to the system, some m/b's "reorder" the boot drive sequence so that the Flash drive isn't tried first. This requires the user to hook up a keyboard and VGA monitor and go into the bios. (But not all m/b's have this problem).

Some m/b's had unreliable drivers for the on-board GigE port.

So we made a decision as a company to offer a pre-built product first. This way we could have a controlled platform and make the system extremely reliable.

Having said this, we are considering publishing a "reference design" which lists the key parts required for reliable operation, and making the software available separately.

I would definitely be interested in a standalone software solution. The number one advantage I see in the Un-Raid is the ability to mix and match drives and upgrade drives one at a time. TO me this sounds like a great backup server. I have an 3ware 8 IDE port card and an Areca PCI-e 1260 16port SATA Raid 6 card as well. the important stuff is duplicated across the 2 arrays and really really really important stuff is archived to optical and tape and stored offsite.

galileo2000

09-01-05, 08:53 PM

I think you missed a decimal place :)... more like 1.6TB. DVD changers are fine as long as they work and you can deal with the wait time between switching disks.

You are correct, I missed the decima, it is 1.6 TB.

Hard drives are also fine as long as they work :D I've seen much less failed DVD changers than I've seen failed HDD's.

galileo2000

09-01-05, 08:58 PM

Galileo,

"I also spent countless hours ripping my CD's. I then reencoded ALL of them with DVD Architect into AC-3 ( and boy, was it a piece of work!). Now I have about 300 CD's stored on 6 DVD's with visual controls and access to any particular CD from the on-screen picture. And guess what, those 6 DVDs are in Sony and Sony plays them"

That sounds like a cool idea! I have never heard of anyone doing it (may have missed it on one of the boards here). Looks like I need to check out DVD Architect and see what that's like. Does the size remain the same when reencoded to AC-3? Do they play in any standard DVD player? Do you have a portable MP3 player and if so, do/did you just also encode to MP3 as well?

No, the size does not remain the same. You can put about 40 to 50 CD's on one DVD. Yes, they play in any DVD player, because player "thinks" it is Video.
Yes, I also encode to MP3 so I can have it anywhere.

If you decide to do what I did with DVD architect, let me know, I'll get you started. I have to tell you it is tremendous amount of work just because I have so many CD's and you need a lot of CPU time to reencode. But the end result is very well worth it. I now know exactly where are my CD's. I started to lose track before.

Sorry for hijacking this thread, did not mean it, just happened.

reel_fan

09-01-05, 11:46 PM

If I started my system with a 300GB parity disk, then decided to move to a 500GB parity disk, can this be done?

Thanks.

Robert Simandl

09-02-05, 01:02 AM

limetech

09-02-05, 02:28 AM

Limetech,

I would definitely be interested in just your software. I already have two CoolerMaster Stacker file servers with eight drives in mobile docks running under Windows XP Pro so I do not need the hardware. However, I believe that your operating system is superior to using Windows XP Pro and would like to encourage you to make it available to AVS forum members who like to "roll their own" like I do.
Regards,
TCIII

We are going to make the s/w available separately, qualified on two different motherboards (plus more if we get some volunteers to help).

It's going to take some time for us to write up the documentation and clear out some previous committments.

limetech

09-02-05, 02:52 AM

I'm pretty cheap and don't find the price to be unreasonable. I don't like that it seems that you can't have a 'pool' of space...I want all my stuff available through a single mount with subdirs and all the space (or however I chop it up) available without running into a situation where my 'movie' volume has 2gb too little but my 'music' volume has plenty of space. I may just be reading the spec and looking at that picture wrong.

Ok, there is a definite advantage to having a uniform name space (or file system). But there are some "technical difficulties" in doing this when file system size approaches 2TB, which you can do with just 5 drives. Not impossible, but will take us some work (need to update our driver for latest linux 2.6.x kernel).

I also fear that the performance from that Celery and its onboard gigE will leave a lot to be desired. All in all though, pretty interesting!
-Trouble

The "Celery" :rolleyes: is plenty fast enough, but we are constantly finding ways to tweak the gigE port for better performance. We also once installed a PCI gigE card and was able to run both it and the on-board gigE at >25MB/s each for reads. Also, surprisingly, we haven't found "jumbo frames" to help much.

We use an Intel motherboard which is based on the 865G chipset. One thing nice about this chipset is that there is a "Dedicated Network Bus" on the northbridge chip which provides a direct data path for the gigE chip straight to memory, bypassing the PCI bus (where all the disk controllers sit). So network activity does not contend with disk activity.

limetech

09-02-05, 03:02 AM

...A couple questions:

1. How do you handle rebuilding? With different size drives, do you have to replace with the same size drive to rebuild, a drive large enough to hold whatever was on the dead drive, or larger drive?

You have to replace with a drive as large or larger than the dead drive. And this presents a problem: suppose your replacement drive is also larger than the parity drive?

In this case we suport a special operation called "Parity Swap". You take your parity drive and use it to replace your dead drive; and you put your new drive in the parity slot. The system will recognize this configuration change and first copy the parity information from your old parity disk to your new parity disk; and, then do a normal rebuild (and expand the file system if possible).

2. Do you support having a backup drive in the system ready to be swapped in as a replacement. If so how does question 1 affect the choice of this drive?

The current s/w does not support hot spare, but this feature is "on the list" :)

limetech

09-02-05, 03:05 AM

how about a Rack mounted version????

or i can roll my own if you do offer the software only option

We've received several requests for a rack mount version. As soon as we find a suitable case, we'll do it.

limetech

09-02-05, 03:10 AM

I would definitely be interested in a standalone software solution. The number one advantage I see in the Un-Raid is the ability to mix and match drives and upgrade drives one at a time. TO me this sounds like a great backup server. I have an 3ware 8 IDE port card and an Areca PCI-e 1260 16port SATA Raid 6 card as well. the important stuff is duplicated across the 2 arrays and really really really important stuff is archived to optical and tape and stored offsite.

You hit it exactly. Also the ability to spin down drives not in use is nice too.

limetech

09-02-05, 03:11 AM

If I started my system with a 300GB parity disk, then decided to move to a 500GB parity disk, can this be done?

Thanks.

Yes. Any drive, including parity can be upgraded to a larger size.

limetech

09-02-05, 03:23 AM

My current server has four drives, almost full, formatted as NTFS. If I were to decommission that server and install those drives into the un-Raid, would the un-Raid be able to recognize/read/stream the data from them? Or will I have to use new blank drives and copy the data over to them after initial setup?

Thanks....

The un-Raid will not recognize the NTFS file system, but will let you re-format them to the ReiserFS (3.6) file sytem.

A word about file systems. We originally wanted to use NTFS because we envisioned the possibility of moving disks between the server and individual windows machines. Though NTFS support in Linux is limited to read-only, there is a company that makes a "NTFS for Linux" commercial product. We abandoned this though because the liscensing was too expensive and besides it was dog slow.

Regarding ReiserFS: for those interested there is a lot of info out there you can google. There is also a free utility you can download that lets you read ReiserFS data under Windows. This is nice to have in case you need to get at some data on the Un-Raid and it's unusable for some reason.

MrSensitive

09-02-05, 09:04 AM

Hi,

I am also very interested in a software only solution,
tha fact that drives can spin down when not in use is a very good power saving option..

would it be able to build a server using a Via mini-itx motherboard having only one PCI slot?

regards

Stereodude

09-02-05, 09:10 AM

Yep, that's a pretty powerful card. I couldn't tell from the documentation, but you probably have to use the same size hard drives to expand with (or larger, but not be able to use the storage until you can create a separate array out of it).
From first hand user accounts on the web I've found, it will expand existing arrays on the fly (which can be over 2TB). They did say it may take 12-24 hours to expand an array though.

Thomas J. Coyle

09-02-05, 10:35 AM

limetech,

I would be glad to test your operating system on the following motherboard: ASUS A7N8X-E Deluxe Socket A (Socket 462) NVIDIA nForce2 Ultra 400 ATX. I presently have this motherbooard installed in my two CoolerMaster Stacker DVD servers. I am running Windows XP Pro on both servers, but have not ripped any DVDs to the second server. Also I would be willing to increase the second server drive capacity from 8 to 12. The Asus motherboard has two NIC ports. One is 10/100 and the other is a Gigabit (Marvell 88E8001) port.
Regards,
TCIII

cburbs

09-02-05, 01:28 PM

What will be the minimum requirements for this software to run?

limetech

09-03-05, 04:58 AM

...
would it be able to build a server using a Via mini-itx motherboard having only one PCI slot?

Do you mean something like this (http://www.viavpsd.com/product/epia_m_spec.jsp?motherboardId=81) ? You'd be limited to 8 hard drives and 10/100 ethernet. I don't know if linux runs on that processor (probably does); so I guess the answer is a definite maybe :)

limetech

09-03-05, 05:02 AM

From first hand user accounts on the web I've found, it will expand existing arrays on the fly (which can be over 2TB). They did say it may take 12-24 hours to expand an array though.

UnRaid takes about 4-5 hours to rebuild/expand a 400GB drive (system can be used during this time though).

limetech

09-03-05, 05:26 AM

What will be the minimum requirements for this software to run?

That is a very good question. We haven't really "scaled back" on the hardware to see where things start to slow down. Our design goals were to support gigEthernet and up to 12 disks. I tend to think the most important factor is motherboard selection. This is because most m/b's put their on-board gigE on the PCI bus where it will share the available bandwidth with the disk controllers. We specifically chose an Intel m/b which has a separate data path to memory for it's on-board gigE (I talked about this in a previous post).

Now in normal use, even with gigE on PCI along with all the disks, this does not necessarily present a problem because you are usually only accessing one disk (reading) or two disks (writing). But we also wanted to be able to sustain high read rates even with a disk disabled in order to support some applications such as streaming video. In this case, all the disks will be read in order to generate the failed disks contents "on the fly", which is going to saturate the PCI bus.

So I think m/b selection should be carefully considered depending your application requirements. Beyond that a "celery" is just fine, though the faster the front-side bus, the better.

Abdul Jalib

09-03-05, 06:34 AM

Ok, there is a definite advantage to having a uniform name space (or file system). But there are some "technical difficulties" in doing this when file system size approaches 2TB, which you can do with just 5 drives. Not impossible, but will take us some work (need to update our driver for latest linux 2.6.x kernel).

A few weeks ago in a RAID thread I suggested that somebody could make a lot of money by making a simple piece of software that just spans together hard drives nondestructively, incrementally, and semi-robustly under Windows. By "semi-robustly," I mean if one drive fails, then the rest of the drives' data can still be recovered (but not that one drive's data.) Windows XP/2000 has spanning, but it's not nondestructive, so I don't think you can easily add and remove drives. Somebody replied to my suggestion, saying there would be a problem with recovering the index if one drive fails, but I think this can be engineered around.

Limetech, you might consider writing such a thing for Linux, and then you can slap it on top of your product, and then consider selling this piece separately as well, especially if you write a Windows version. It might compete with your hardware product, or it might bring a lot of people to your website and sell your hardware product.

J. L.

09-04-05, 11:04 AM

UnRaid takes about 4-5 hours to rebuild/expand a 400GB drive (system can be used during this time though).
Your solution for network storage looks very attractive... I too like the idea that it can be expanded incrementally as disks get cheaper and space is filled. I like that it can be used for all bulk storage in my home lan and that it will spin down most of the drives when not accessed for a period of time. I like the idea of a wake-on-lan solution since I currently use that feature to remotely enable a server in my basement today.

Could you answer a few questions that I did not find in your FAQ:

1) Since the NAS storage continues to be available when a drive fails, and we all know that eventually a drive will fail, is there any external indication of the failed drive? Does the OS send an e-mail, flash lights, sound an alarm or do something to let me know I need to attend to the failed drive before a second disk fails? In my mind it is not sufficient for the status to be only available on the administrative screen on the NAS server since that screen is likely to never be visited frequently once initial configuration is done, especially since there is no other outward indication a drive is in trouble (the recreated data from the failed drive is still available, so we will not notice when a single drive fails).

2) Can you share your ideas on how software updates will be handled/made available to early customers? (you mentioned an upgrade to the latest Linux kernel and possible changes to your raid driver to allow a consolidated name space on a single "network share" spanning multiple physical drives as two potential upgrades)

3) Are there any plans to add to your OS software that can monitor an attached UPS? Since it is very likely that some of us will use a UPS, can the UPS signal (through the serial port perhaps) the OS to perform an orderly shutdown when its batteries are nearing depletion in a power outage? Yes, I know this is very advanced for most users and that the software to monitor the UPS is likely brand specific, but it would be nice if it was possible.

Joe L.

limetech

09-05-05, 04:02 AM

A few weeks ago in a RAID thread I suggested that somebody could make a lot of money by making a simple piece of software that just spans together hard drives nondestructively, incrementally, and semi-robustly under Windows. By "semi-robustly," I mean if one drive fails, then the rest of the drives' data can still be recovered (but not that one drive's data.) Windows XP/2000 has spanning, but it's not nondestructive, so I don't think you can easily add and remove drives. Somebody replied to my suggestion, saying there would be a problem with recovering the index if one drive fails, but I think this can be engineered around.

Limetech, you might consider writing such a thing for Linux, and then you can slap it on top of your product, and then consider selling this piece separately as well, especially if you write a Windows version. It might compete with your hardware product, or it might bring a lot of people to your website and sell your hardware product.

First the "easy" way. At some point we will switch to the linux 2.6.x kernel (using 2.4.x now). This will let us "span" the storage of all installed hard drives into one large volume >2TB (if so desired). You could still add additional disks to expand storage, but you would not be able to expand an existing disk by replacing it with a bigger one.

The "hard" way would be to implement some kind of layer as you describe that would be able to merge directory trees of individual drives into one "view". Along with that would be code that heuristically decided which hard drive to use when a new file is created (such as just pick the one with the most free space). Anyway, we've given this some thought and perhaps it's feasible.

limetech

09-05-05, 04:21 AM

1) Since the NAS storage continues to be available when a drive fails, and we all know that eventually a drive will fail, is there any external indication of the failed drive? Does the OS send an e-mail, flash lights, sound an alarm or do something to let me know I need to attend to the failed drive before a second disk fails? In my mind it is not sufficient for the status to be only available on the administrative screen on the NAS server since that screen is likely to never be visited frequently once initial configuration is done, especially since there is no other outward indication a drive is in trouble (the recreated data from the failed drive is still available, so we will not notice when a single drive fails).

At the present time the only indication of disk failure is on the browser main page; which means you have to look at it once in a while. Having said that, all the code is already "under the hood" to be able to send email alerts. This feature is near the top of the list for the next software release.

2) Can you share your ideas on how software updates will be handled/made available to early customers? (you mentioned an upgrade to the latest Linux kernel and possible changes to your raid driver to allow a consolidated name space on a single "network share" spanning multiple physical drives as two potential upgrades)

Here's how software update works. Besides the disk shares, the unraid server exports one more share called "flash". This share gives you access to the contents of the flash drive. To upgrade the software you download the release from our website (a zip file). You then unzip the contents which consists of a directory named "current" which contains the files of the new release. You then drag that directory to the "system" directory on the flash share. Finally, you go to the management utility and restart the system.

At some point we'll make a small windows utility that does all this automatically. Also, you can do the same thing by plugging the Flash drive into a PC and copying the files directly to it.

3) Are there any plans to add to your OS software that can monitor an attached UPS? Since it is very likely that some of us will use a UPS, can the UPS signal (through the serial port perhaps) the OS to perform an orderly shutdown when its batteries are nearing depletion in a power outage? Yes, I know this is very advanced for most users and that the software to monitor the UPS is likely brand specific, but it would be nice if it was possible.

Yes, UPS support is part of our package that includes shutting down into standby mode after a (configurable) period of inactivity. The UPS control/monitor s/w is brand specific, but most manufactuers provide the s/w we need.

We'll be updating our website soon with our development road map along with access to the download area.

timexx8

09-05-05, 06:21 AM

We've received several requests for a rack mount version. As soon as we find a suitable case, we'll do it.

I'm ready to purchase when you offer a rack mount version, sata preferred.

Abdul Jalib

09-05-05, 08:41 AM

I'm ready to purchase when you have a European distributor.

J. L.

09-05-05, 10:14 AM

At the present time the only indication of disk failure is on the browser main page; which means you have to look at it once in a while. Having said that, all the code is already "under the hood" to be able to send email alerts. This feature is near the top of the list for the next software release.Although I do not know what is currently at the "top of the list" (bug fixes?) this should be bumped up in priority. Once in place, your server/NAS will be forgotten until it is not accessable... and that will be too late if multiple disk failures was the cause. An e-mail alert, in addition to a text alert sent across the LAN network to "Yak" clients, or something like them (http://sunflowerhead.com/software/yac/) running on PCs on the LAN would go a long way to having a proactive response to a failure rather than a reactive one.

I use something like it to display caller-ID on all the PCs on my LAN today, adding a server disk-failure alert could be straightforward. (a small PC program, perhaps even "yak" on clients, and a hook into your disk management to send the udp packet to clients listening on the lan might not be too hard to impliment)

Another idea is a small windows-based program that can query the health of the NAS server and report if something needs attention. This can be run on start-up on the windows PCs, or run on a schedule if they manage to stay up without rebooting, to let us know if the server needs tending to.

the unraid server exports one more share called "flash"
Nice that upgrades are easy to perform. That share should not be open to writing by all by default. Perhaps it is password protected against accidental damage.

limetech

09-07-05, 05:43 AM

Although I do not know what is currently at the "top of the list" (bug fixes?) this should be bumped up in priority...

Top 3 features for next release:
1. Security
2. Alerts
3. Standby ability

...An e-mail alert, in addition to a text alert sent across the LAN network to "Yak" clients, or something like them running on PCs on the LAN would go a long way to having a proactive response to a failure rather than a reactive one.

Your idea of using a Yak client is very interesting; we'll definately look into that. Also, if anyone reading this is a girder expert: can girder be configured to respond to a "magic packet" sent via configurable tcp/udp port?

...Nice that upgrades are easy to perform. That share should not be open to writing by all by default. Perhaps it is password protected against accidental damage.
Protection of this share is covered under the "Security" feature.

Thanks for the great ideas - keep them coming :)

musicmann

09-07-05, 09:45 AM

limetech

09-08-05, 02:21 AM

Limetech, what's the timeframe for:
1. Version 2 of the software
2. Documentation et al needed for releasing a software-only product

I'm pretty excited about this product...I just have to figure out how to quickly use up my remaining 800 gigs...maybe I'll take some days off of work to finish ripping my DVD collection...
Our plan at this point is to introduce the software-only product before any major new features, probably near the end of this month. We are qualifying two more m/b's at present along with clearing some previous commitments. So we'll see who can finish first: us releasing the s/w-only product or you ripping 800GB of media :D

albrigsr

09-08-05, 10:22 PM

I received my order today. It comes in 2 separate packages, the flash card and the server unit. I ordered it on September 2nd so total turnaround time was 1 week (as fast I could hope for!). The packaging was great...that spray in foam stuff. The CoolerMaster case is a monster. It looks like there is plenty of room to keep air moving and the case should be well-ventilated. It seemed extremely solid including the removable drive cases. The wheels on the bottom of the case are a HUGE help. That's a great design feature as I own another server case that doesn't have them and it is a real pain to move around. The wheels rolled well on both the hardwoods and carpet in my house.

Anyway, I can't do anything except look at it until my 2 500GB drives come which should hopefully be tomorrow or Saturday. I will follow-up with my experience after the initial installation.

In case anyone didn't read my earlier post my plan is to hook up the 2 500GB drives and copy the entire contents from an existing windows formatted drive on my network onto the 500GB drive. I will then move the drive that I copied from into the server where it will be re-formatted to the ReiserFS format. I will then repeat the process until I have moved all the excess storage drives from each computer on my network into the server. I will try to keep some benchmark times for each activity but can't promise I will pay too close attention due to 2 small children demanding more of my attention! I think I will end up with the following in the server:

1 - 500GB Parity Drive
1 - 500GB Data Drive
2 - 400GB Data Drives
1 - 300GB Data Drives
2 - 250GB Data Drive
1 - 120GB Data Drive

Total - Roughly 2 TB of Data

That would still leave me with 3 open drive slots and I will eventually upgrade the smaller drives (250GB, 120GB) as needed with larger drives like the 500GB.

reel_fan

09-09-05, 12:07 AM

For those of us who work practically next door, do you have a will-call? If so, is there a discount :rolleyes:

limetech

09-09-05, 02:25 AM

I received my order today. It comes in 2 separate packages, the flash card and the server unit. I ordered it on September 2nd so total turnaround time was 1 week (as fast I could hope for!). The packaging was great...that spray in foam stuff. The CoolerMaster case is a monster. It looks like there is plenty of room to keep air moving and the case should be well-ventilated. It seemed extremely solid including the removable drive cases. The wheels on the bottom of the case are a HUGE help. That's a great design feature as I own another server case that doesn't have them and it is a real pain to move around. The wheels rolled well on both the hardwoods and carpet in my house.

Yes, "solid" is a good description. We spend a lot of time on component selection.

Anyway, I can't do anything except look at it until my 2 500GB drives come which should hopefully be tomorrow or Saturday. I will follow-up with my experience after the initial installation.

You could install the Flash drive and power up the unit. After it has completed boot process, you should see the system show up in My Network Places with a single share called "flash", which is the contents of the flash drive. You can do this to verify your network connection. (If you don't use DHCP there are instructions on our website for setting up a static IP address.)

In case anyone didn't read my earlier post my plan is to hook up the 2 500GB drives and copy the entire contents from an existing windows formatted drive on my network onto the 500GB drive. I will then move the drive that I copied from into the server where it will be re-formatted to the ReiserFS format. I will then repeat the process until I have moved all the excess storage drives from each computer on my network into the server. I will try to keep some benchmark times for each activity but can't promise I will pay too close attention due to 2 small children demanding more of my attention! I think I will end up with the following in the server:

1 - 500GB Parity Drive
1 - 500GB Data Drive
2 - 400GB Data Drives
1 - 300GB Data Drives
2 - 250GB Data Drive
1 - 120GB Data Drive

Total - Roughly 2 TB of Data

That's a good strategy.

That would still leave me with 3 open drive slots and I will eventually upgrade the smaller drives (250GB, 120GB) as needed with larger drives like the 500GB.
I count 4 open drive slots....

limetech

09-09-05, 02:30 AM

For those of us who work practically next door, do you have a will-call? If so, is there a discount :rolleyes:

Unfortunately we don't offer will-call at this time. We're currently looking for a larger facility which would lets us do this. (It shouldn't take too long considering the local economy here in Silicon Valley :rolleyes: )

madpoet

09-09-05, 09:34 AM

Can you explain how the Raid4 works? Does it have an entirely seperate parity drive that maintains the parity for the other 11? Looking forward to the seperate software/hardware list version!

albrigsr

09-09-05, 03:53 PM

4 open drive slots is correct! (and I have a Master's degree in accounting!).

I thought I could boot it up w/o any drives but I thought the directions said to initially install the drives first. Anyway, I will go ahead and insert the flash and see if I find it on my network if I don't get the drives this weekend.

I could take some pictures if any wants but nothing is different than what is on the website.

limetech

09-10-05, 02:51 AM

Can you explain how the Raid4 works? Does it have an entirely seperate parity drive that maintains the parity for the other 11? Looking forward to the seperate software/hardware list version!

Yes, there is a separate parity drive that is updated any time you write one of the data drives. We chose to designate the top hard disk tray (or slot) to be the parity drive.

We call our system an "unRaid" because, while similar to Raid4, there is no "RAID" designation for the organization we're using.

[Why do we say "un"-RAID - well it's from the old 7-up ad campaign of the 70's where they called it the "un-cola". :rolleyes: ]

Anyway, in a true RAID-4 system, data is striped across all the data drives. This means that the aggregate storage of the data drives is considered a single volume. Block 0 of this volume is on block 0 of disk 1; block 1 is on block 0 of disk 2, block 2 is block 0 of disk 3; etc.. Eventually the block numbering wraps back around to disk 1 again. The number of data disks you can write to before the block numbers wrap around is called the stripe size.

The big adavantage of striping is the great magnification of transfer rate you can achieve. If you are reading a large file spanning many blocks, then the drives are going to operate in parallel to supply the data.

Now this scheme was really good when hard drive data transfer rates were 1-5 MB/sec "back in the day". It opened up many possibilities for applications, especially video. For example, Disney's movie "The Little Mermaid" (1989) was largely produced using workstations moving data on and off 4-disk Raid-0 disk arrays.

Today, however, things have changed. A typical 7200 RPM drive can transfer data at 60 MB/sec (though slower as you move toward the inner cylinders). Just 2 of these drives running in parallel uses all the bandwdith of a 33MHz/32-bit PCI bus. So for applications that matter to HTPC and many other streaming media applications, other factors become more important than sheer disk bandwidth.

The unRaid does not stripe the data. When you read a large file which spans multiple blocks, they are all read off of the same drive. But since hard drives are so much faster now (and getting faster), the data rate is plenty fast enough. IMHO, traditional RAID systems will soon be relegated back to where they began: supercomputer memory swap space.

limetech

09-10-05, 03:07 AM

4 open drive slots is correct! (and I have a Master's degree in accounting!).
A Master's degree is no match for 2 small children :D

I thought I could boot it up w/o any drives but I thought the directions said to initially install the drives first. Anyway, I will go ahead and insert the flash and see if I find it on my network if I don't get the drives this weekend.

What we want to try and avoid is to power-up the system without the Flash drive installed. This is because some M/B's, noticing there is no Flash, will change the boot order to, say, the "primary master" which of course won't boot. If this happens you have to go into the bios and fix it. NOTE: this problem does not seem to occur with the latest "P24" version of the bios on the Intel M/B we are using.

madpoet

09-10-05, 09:08 AM

Thanks limetech... like I said, really looking forward to the software version.

mpenton

09-10-05, 10:45 AM

I'd also be interested in the software version. I'm already running R5 under a version of Linux booting from a CF drive, using the Promiss controllers and gEnet. The thing I'm after is the ability to 'painlessly' expand the system without having to perform too many magic tricks. Please let us know when the software is ready.

albrigsr

09-10-05, 12:14 PM

Ok. I received my 2 500GB drives yesterday and installed them today. I have never worked with a system that uses drive caddies and bsolutely love it. Especially because after I installed the 2 drives I realized that I had screwed up the jumper settings! It took me all of about 5 minutes to remove the 2 drives, change the jumper settings and reinstall (that includes removing and reinstalling the screws). The whole unit (at least with just the 2 drives installed) seems very quiet. I didn't turn on both power supplies as I don't need to until I install a drive in one of the bottom 6 slots. I also got to try the re-initiale the array setting as I didn't realize my incorrect jumper setting until after I had formatted the drives and saw just 32GB instead of 500GB. It work easily enough and now the drives are formatting/clearing/syncing or whatever the technical term. I started at 10:35am CST and it reports another 164 minutes with an estimated 41,482 KB/sec. That would make the entire process about 204 minutes long which is close to what it initial predicted. I'll be interested in the initial transfer speeds as I begin to move data over my network (100MB not Gig unfortunately...). I'm not too concerned about write speed as most moves will be post ripping a DVD and done every once and a while but still am interested in how long the entire process of moving things to the server will last. So far I am extremely happy as the unit has been very user-friendly.

madpoet

09-10-05, 04:10 PM

Do you have the SATA or IDE version?

albrigsr

09-10-05, 07:20 PM

IDE. I don't think the SATA version is available yet. Iwent IDE b/c the all of my existing drives are PATA.

3-way

09-10-05, 07:30 PM

I didn't see an SATA version of the Cremax removable bays. Is there another comparable alternative?

CANNON-FODDER

09-10-05, 08:03 PM

http://www.cremax.com/

MB123AK - PATA-PATA
MB123SK - SATA-SATA
MB123PK - PATA HHD - SATA MB

v/r,
C-F

limetech

09-11-05, 02:14 AM

http://www.cremax.com/

MB123AK - PATA-PATA
MB123SK - SATA-SATA
MB123PK - PATA HHD - SATA MB

v/r,
C-F

Our SATA prototype uses the MB123SK mobile rack. So far we have found Cremax to be the best value and very solid; and we have tested A LOT of mobile racks.

justmikie

09-12-05, 01:24 PM

Since I have a few SATA drives that I would like to be able to stick in when I place my order, has anybody tried any of the IDE to SATA converters? I found a few places that make them, but don't have any experience with them. Do they work? Do they slow down transfer rates noticeably/a lot/not at all?

starshipwarrior

09-13-05, 12:38 AM

Well you got my attention - questions
1. What is the fastest CPU that this mother board will take ?
2. What is the max RAM this MB can handle ?
3. How much trouble would it be to put in twin 350s PSU / heat?

I do not have a network - so I have no experience with a server and setting one up - I know all the facts of a network just never had any experience with one - and now as times change and work load increases its time to get on the stick

I currently use 5 systems/ All using XP Pro - Each system has one internal SCSI HD plus 7 SCSI HDs in external boxes I work in HD video now !

The plan would be to set up a network and use this server (I'm thinking rack now that you plan to offer it soon) with the 5 systems, I use ASUS motherboards – I do not want to transfer everything I have to the server but use it as addition storage and as the media server

The server would have to handle the following concurrently
1- provide the movies for the HTPC
2. provide the video for what ever I am working on (a lot of HD video files)
3. provide direct storage for downloads off the Internet
4. provide direct storage for TV recording
5.provide data for burning DVDs
6.provide data output to kitchen system for our receipt files

As you can see this machine will be working very hard at the very minimum three people will have to access it at the same time
example I would need to do "1,2, and 5 at the same time or 1,2 and 4
How many of these can the server handle at once - and not trip over itself
Am I barking up the wrong tree here ?

Thank You for your time and efforts

Two Cents

Starship Warrior

limetech

09-13-05, 04:04 AM

You have quite a list here...

Well you got my attention - questions
1. What is the fastest CPU that this mother board will take ?
2. What is the max RAM this MB can handle ?
3. How much trouble would it be to put in twin 350s PSU / heat?

The board can accept up to a 3.4GHz Pentium 4 and up to 4GB of memory... However, in it's current form, there's not a lot of benefit using that fast of a processor and that much memory because:
a) The UnRaid software is not cpu-bound, it's "PCI bus" and "Disk I/O" bound.
b) Having more memory would let the system cache more data, which would speed up throughput if your I/O stream tends to be "bursty". However, in video applications you are probably going to be streaming more data than can be held in the cache so you'll be ultimately bound by the disk I/O anyway. Perhaps this is something we could experiment with though.

Regarding dual 350 PSU's: yes that should not present any problem. We chose our PSU's very carefully. The ones we use provide 18A on the +12 rails. +12 is what is used the most by hard drives (esp. Seagates).

I do not have a network - so I have no experience with a server and setting one up - I know all the facts of a network just never had any experience with one - and now as times change and work load increases its time to get on the stick

I currently use 5 systems/ All using XP Pro - Each system has one internal SCSI HD plus 7 SCSI HDs in external boxes I work in HD video now !

Yikes! You have 5 systems and they're not networked together? :eek:
Windows networking has come a long way since Win95 days; XP Pro networking is very easy to set up.

The plan would be to set up a network and use this server (I'm thinking rack now that you plan to offer it soon) with the 5 systems, I use ASUS motherboards – I do not want to transfer everything I have to the server but use it as addition storage and as the media server

A good plan.

The server would have to handle the following concurrently
1- provide the movies for the HTPC
2. provide the video for what ever I am working on (a lot of HD video files)
3. provide direct storage for downloads off the Internet
4. provide direct storage for TV recording
5.provide data for burning DVDs
6.provide data output to kitchen system for our receipt files

1- yes
2- yes - If I may ask: what is the source of your HD video?
3- yes
4- yes, Is this HD as well? How many simultaneous write streams?
5- yes
6- yes (I guess you mean recipe)

As you can see this machine will be working very hard at the very minimum three people will have to access it at the same time
example I would need to do "1,2, and 5 at the same time or 1,2 and 4

The most critical item in your list is number 1 (viewing movies) because of the fact that if this data stream is interrupted, then you'll see "jerkiness" in what you're viewing. We haven't extensively loaded the machine to see if we could disrupt a video stream. But we have done this: with a disk disabled, view a movie stored on that disk, while at the same time view a movie on another disk. We've also viewed a move while at the same time creating a video file. None of these operations caused any drop outs (using Zoomplayer). Ok, we have some more load testing to do...

How many of these can the server handle at once - and not trip over itself

There are two things that limit the performance of the server:
1. Write performance. Since there's a single parity drive, you are going to get somewhere around 12 Mega-Bytes/sec. [though we are starting to improve this somewhat]
2. Network performance. The GigEthernet port seems to peak out around 25MB/sec, though we're not exactly sure why this happens. I would expect to be able to improve this further.

Am I barking up the wrong tree here ?

For the applications you cited, these performance numbers should be adequete.

albrigsr

09-14-05, 11:01 PM

I just wanted to provide everyone an update. I am currently moving my 5th disk onto the server. Everything has been working extremely well and adding disks is a breeze. My biggest issue is that I currently have a 100MB/s network rather than GigE and it takes a LONG time to move 400GB off a computer to the server. I plan on eventually moving to GigE to take advantage of the faster connection but currently only the server has such a card/connection. The disks all run very cool thanks to the fans and the case. I certainly would recommend the product to anyone at this stage. I have been able to hook my xbox up to it and watch movies. I haven't had the time to stress test it by watching 4 DVD streams at once (which I can try although can't think of it happening at my house....maybe 1 or 2) but it would probably be limited by my network before the server.

Anyway, AWESOME product. Thanks Tom and Lime-Technology!

limetech

09-15-05, 04:20 AM

I just wanted to provide everyone an update. I am currently moving my 5th disk onto the server. Everything has been working extremely well and adding disks is a breeze.
...
Anyway, AWESOME product. Thanks Tom and Lime-Technology!

I'm glad you're happy with the product and thanks for the kind words :)
(And no, albrigsr is NOT my brother-in-law :D )

Thomas J. Coyle

09-15-05, 03:14 PM

limetech,

Are you still on track to deliver a s/w only operating system package by the end of September? If so, can you list the motherboards that you plan to support?
Regards,
TCIII

starshipwarrior

09-15-05, 06:07 PM

Tom,
I sent an email to the company (due to its lenght ) did you get if if now please

pvt me

Starship Warrior

limetech

09-16-05, 01:26 AM

limetech,

Are you still on track to deliver a s/w only operating system package by the end of September?

Yes.

If so, can you list the motherboards that you plan to support?
Regards,
TCIII

We'll publish the initial list of m/b's at the time we make the s/w available. The m/b we currently use is the Intel D865GLCLK.

Thomas J. Coyle

09-16-05, 09:01 AM

limetech,

Thanks for the update. Looking forward to the s/w package release!
Regards,
TCIII

J. L.

09-16-05, 03:12 PM

Tom@limetech,

I am seriously considering ordering the full "hardware solution" from you rather than assembling my own case,drive racks, etc. I know you just said in a prior post the OS-flash drive would be available as a stand-alone product soon, but I need the hardware anyway and getting everything assembled is worth a few dollars more expense to me.

I do have a few more questions. With any luck, some others might also be interested in your answers, so I'm doing this in the thread rather than via PM.

If power is lost and then re-applied some time later, does the system re-start itself? or is user interaction required?

You mentioned in passing on your web-site that "experienced users" can add their own "packages" in the start-up script on the flash drive. To me that is a great feature, especially since I am one of those "experienced users" comfortable in the UNIX/Linux environment (Been working in UNIX since 1980 and have written UNIX device drivers used in commercial products... Back in the early days of Linux, version 0.98 I think, I re-wrote the device driver for my sound-card/scsi controller so it would properly initialize the sound card and then and re-compiled a custom kernel under the Slackware release)

Does your loader/initialization script support loading "modules" for other hardware/devices not currently in your kernel?

What "distribution/release" of Linux is your flash-drive-version based on? (How easy is it to get compatible "packages" for us "experienced users")

When you use the Intel D865GLCLK motherboard to configure a server are there any open PCI slots or are they all occupied?

Most of these questions are somewhat related. I would love to use the server in a home-automation/security role when we are away from the house. (during that time, it would not be serving up media and largely idle... perfect for my use in the alternate role.)

Although this next question comes under the category of "very very experienced users" are there tools on the flash drive to recompile/relink the kernel? If not using the tools on the flash drive, can it be done by us "very experienced users" using a machine loaded with the development tools (compiler, etc) Can you supply the .o for your custom device driver so it can be linked? or is it a module so it just needs to be loaded?

Is there a minimal "shell" on the flash drive? how about X-windows? Perl?

Can we "telnet" in and get a command prompt?

Lastly, since I've already had my credit card info stolen once from a compromised web-based ordering system and had to get a new card issued, is it possible to place an order via telephone rather than via the web? I'll be happy to call you (send your number via PM if you wish and include when it is appropriate to call... or I can give you my number and it will be your dime.)

Joe L.

limetech

09-17-05, 02:19 AM

...
If power is lost and then re-applied some time later, does the system re-start itself? or is user interaction required?

Yes the system will automatically restart. User intervention is required only if there's been a disk configuration change.

You mentioned in passing on your web-site that "experienced users" can add their own "packages" in the start-up script on the flash drive. To me that is a great feature, especially since I am one of those "experienced users" comfortable in the UNIX/Linux environment (Been working in UNIX since 1980 and have written UNIX device drivers used in commercial products... Back in the early days of Linux, version 0.98 I think, I re-wrote the device driver for my sound-card/scsi controller so it would properly initialize the sound card and then and re-compiled a custom kernel under the Slackware release)

Does your loader/initialization script support loading "modules" for other hardware/devices not currently in your kernel?

What "distribution/release" of Linux is your flash-drive-version based on? (How easy is it to get compatible "packages" for us "experienced users")

Ok, here are some linux details. We started with Slackware 10.1 distro, and then pruned that down to about 60MB; further pruning is definately possible. During operation the entire root directory tree is mounted on a ram drive initialized at boot time from a file stored on the Flash (so-called "initrd").

We are able to store two complete "versions" of the software on the Flash: the current running version and a back up copy. On the 128MB Flash, we have about 55MB free. Since the kernel supports loadable modules we definitately anticipate support for other types of devices; currently the unRaid driver and all the ethernet drivers are modules.

When you use the Intel D865GLCLK motherboard to configure a server are there any open PCI slots or are they all occupied?

This particular m/b has 3 PCI slots and we are using 2. We will probably offer an option to install another GigEthernet PCI port. Of course there are plenty of USB ports from which any number of different devices could be attached. Finally there's a IEEE1394 (aka, Firewire) port. Some day we hope to use this as a host attach.

Most of these questions are somewhat related. I would love to use the server in a home-automation/security role when we are away from the house. (during that time, it would not be serving up media and largely idle... perfect for my use in the alternate role.)

At the present time we are taking a conservative approach to opening up the system for add-on packages. The most important role for our system is protecting the data stored on it. As you add more "stuff" to it, the chance of running out of memory or a system crash is much much higher which could result in user data loss.

Although this next question comes under the category of "very very experienced users" are there tools on the flash drive to recompile/relink the kernel? If not using the tools on the flash drive, can it be done by us "very experienced users" using a machine loaded with the development tools (compiler, etc) Can you supply the .o for your custom device driver so it can be linked? or is it a module so it just needs to be loaded?

There's not enough space on the Flash for the full kernel sources and build environment, but you could certainly build modules on a "development" system and have them loaded at boot time on the unRaid server. As explained above, the drivers are modules so no need to supply object files.

Is there a minimal "shell" on the flash drive? how about X-windows? Perl?

Can we "telnet" in and get a command prompt?

The full "bash" shell is present, no X-windows, no Perl (though that could easily be added). You can telnet to the box and login under "root".

Lastly, since I've already had my credit card info stolen once from a compromised web-based ordering system and had to get a new card issued, is it possible to place an order via telephone rather than via the web? I'll be happy to call you (send your number via PM if you wish and include when it is appropriate to call... or I can give you my number and it will be your dime.)

Yes, we can accept different forms of payment. We'll get the order system updated this weekend in order to do this.

Once again, thanks for all the great questions!

albrigsr

09-17-05, 02:51 AM

limetech

09-17-05, 03:23 AM

Tom,

I am sitting in the Zurich airport waiting on a connecting flight and just saw your message regarding the USB connections on the server. I was wondering on the trip over if I could hook up an external Hard Drive (USB) to the server and copy data to/from that drive. I would try it if I was at home but won't be for another 10 days so the question is "Will it work"? If so, are the USB connections USB2.0 or 1.1? I guess it wouldn't be worth it if they were only 1.1.

You know that's a good idea... I haven't actually tried it, but I think there's no reason why it shouldn't work as long as linux can mount the file system. Also there's no user i/f for it in management utility, so you'd have to use a command line interface to do it, for example, telnet into the box or hook up a monitor and keyboard. The ports are USB2.0.

Also, I saw somewhere that you were getting 25MB/s over GigE. Is that correct? I assume that is MB/s and not Mb/s, right? I am asking as I am wondering how much of a real increase in speed would I see in moving from 100 Mb/s network compared to a Gig network. I have heard 2-3 times faster as a rule of thumb. Does this sound right or are there some other actual limitations (I know I have read disk/bus speed truly constrain a Gig network)?

Yes we see around 25,000,000 bytes/second for reads. So this is around 3x what you can get with "Fast" ethernet (100Mbit). We're actually spending a great deal of time investigating just where the bottlenecks are which limit GigE performance; and there are alot of places to look:
- The linux ethernet driver stack
- The windows ethernet driver stack
- The network itself, ie, cabling, switches, etc.
In addition, each driver stack typically has several tunable settings, not to mention the TCP/IP specific tunable settings such as frame size, window sizes, etc. etc. If we experience any "breakthoughs" I'll be sure to post them.

But even with "only" a 3x increase, going to GigE is definitely worth it.

J. L.

09-17-05, 05:46 AM

Yes the system will automatically restart. User intervention is required only if there's been a disk configuration change.Nice

Ok, here are some linux details. We started with Slackware 10.1 distro, and then pruned that down to about 60MB;Thank you for the details... I figured the 128meg flash drive was too small for the full source&development toolset, but wanted to ask which "release" you were based on in case I wished to add anything.
On the 128MB Flash, we have about 55MB free. Since the kernel supports loadable modules we definitately anticipate support for other types of devices; currently the unRaid driver and all the ethernet drivers are modules.Nice

This particular m/b has 3 PCI slots and we are using 2. We will probably offer an option to install another GigEthernet PCI port. Of course there are plenty of USB ports from which any number of different devices could be attached. Finally there's a IEEE1394 (aka, Firewire) port. Some day we hope to use this as a host attach.Perfect in that a spare slot is available.
At the present time we are taking a conservative approach to opening up the system for add-on packages. The most important role for our system is protecting the data stored on it. As you add more "stuff" to it, the chance of running out of memory or a system crash is much much higher which could result in user data loss.I fully understand the need for being conservative. All it takes is one malloc() combined with a branch through the code where the free() call does not occur to have a "memory leak" that will eventually crash the system. I also fully realize the potential of a unRaid server/processor running 24/7 in my home. I also realize that any add-ons are at *MY* risk and most important to you is the stability of the unRaid driver and associated disk storage.

The full "bash" shell is present, no X-windows, no Perl (though that could easily be added). You can telnet to the box and login under "root".telnet is fine. I realize the distribution is trimmed down to get everything to fit the flash drive. Don't add Perl on my behalf. I was just asking.

As far as my other ideas, I was thinking of the possibilities of running a video capture card in that spare slot and V4L/zoneminder for security. Ideas only at this time, but as a fellow developer, I hope you understand... Since I plan on setting up V4L/zoneminder anyway on a home automation PC and figured I would ask if a slot was free in the unRaid server in case sometime in the future I wanted to try to combine both functions on the same PC and save a bit on my electric bill. I'm fairly sure those developers of V4L are equaly paranoid/cautious about system stability.

Yes, we can accept different forms of payment. We'll get the order system updated this weekend in order to do this.

Once again, thanks for all the great questions!I think you might have misunderstood my question. Payment via credit card is fine, but I prefer to do it via a phone call vs. a web-based form.

Please don't take this wrong, but I am guessing you are a very small company and work out of your home. You probably have a "day job" to support your family and are reluctant to post/give out a personal phone number. I've got no problems with either, but as I said in my prior post I once had my credit card info stolen and held hostage by web-extortionists. I had to get a new card to replace it. That was over a year ago, and no charges were ever made to my old account, but as a result, I prefer to do my ordering via the phone.

You value your privacy, and as a result use a different "handle" for the unRaid server then any of your other posts on AVS or any other forum. (the only posts from "limetech" are in this thread) On the other hand, I too value my privacy... to do business we need to meet in the middle (and not in a public forum) I'll check out your web-site in a few days to see the updates and hope to make contact with you in the next week or so.

I like the idea of the unRaid server and after reading favorable feedback from your unBrotherInLaw :) I am looking forward to having an unRaid server in my home.

Joe L.

Diode1

09-17-05, 10:33 AM

Most credit cards now have a web-based feature that will let you generate a one-time use number. You pick the max amount and the expiration month.
MBNA calls it Shop Safe, I am sure it goes by many names.
You can do this in two manners, log onto your credit card website and use the feature there.
Or dload the small application and just do it directly from within you pc.
This way you never have to give out your actual account number and it will expire at your predetermined time.

Tedd

09-24-05, 09:53 PM

I also am interested in the usb software version as I already have a fair amount of computer hardware on hand for this project and just purchased a D865GLCLK motherboard. (The new desktop has a D865GRH board so I was pretty close there...)

Any idea when we can expect the SATA version? And what controller do I need for SATA?

Govnah

09-25-05, 11:05 AM

Count me in on the software as I have the hardware ready to go......Also any ETA on the software?

TIA

limetech

09-26-05, 03:00 AM

Ken Greene

09-27-05, 07:00 AM

how about a mix of sata and IDE controlers ???? is that possible?

EricD

09-27-05, 10:48 AM

If your OS is on one disk and your "unraid" is software raid, then what happens when you lose that drive?
If your parity drive is only one drive and that fails, then what happens if another drive fails? Do you have automatic failover?

It sounds like the "unraid" is just that, so why even have it?

jimwhite

09-27-05, 11:04 AM

the OS is on a USB flash drive....

even Raid 3,4, and 5 cannot recover from 2 failed drives....

:cool:

EricD

09-27-05, 12:03 PM

If you have configured hot spares they sure can.

So what happens if the flash drive gets messed up? Bye Bye data?

xact

09-27-05, 09:42 PM

If you have configured hot spares they sure can.
A hot standby drive would be a nice feature. Maybe they could add this to the un-RAID software eventually.
So what happens if the flash drive gets messed up? Bye Bye data?I imagine you'd copy their Linux distro to a new $20 flash drive. I don't think it would be a big deal.

limetech

09-28-05, 02:08 AM

how about a mix of sata and IDE controlers ???? is that possible?

Yes, that should be possible. Probably not something we would offer as a preconfigured product, but definitely something you could do once our SATA version is complete (along with a little set up help from Cust. Support).

limetech

09-28-05, 02:26 AM

If your OS is on one disk and your "unraid" is software raid, then what happens when you lose that drive?

The OS is actually running entirely from RAM. At boot time, the compressed image is read from the Flash. Thereafter, the Flash is only accessed to record configuration changes.

If your parity drive is only one drive and that fails, then what happens if another drive fails? Do you have automatic failover?

After any one drive fails, either parity or a data drive, the array is said be be "unprotected". While in this state, if another drive fails, then data may be lost on both drives. I say "may" because of this:
- If one of the failed drives is the parity drive, then of course there's no user data on that drive.
- Depending on the nature of the failure, data may still be recovered. The UnRaid system deems a drive as "failed" when either a) it is not recognizied at boot time, or b) an unrecoverable write error occurs on that drive.

Now in case a), probably you've lost the drive. But in case b) you may be able to still read data from it.

This is one advantage to the UnRaid orgainization. Since data is not striped, you can always just connect any individual drive up to another system "normally" and access the data on it. Granted, we use a linux file system (Reiserfs), so if you want to mount the drive in a Windows machine, you need to download a utility to let Windows recognize the file system.

limetech

09-28-05, 02:45 AM

If you have configured hot spares they sure can.

Yes a hot spare might be a nice future addition. Of course if another drive fails during the resync process to the hot spare, then you've probably lost data. All a hot spare does is decrease the time exposure following a drive failure during which you are vulnerable to another drive failure.

We have not implemented the hot spare feature immediately because:
a) we found that people tend to buy drives on an "as needed" basis and/or when a good deal comes along (e.g., a good rebate). Since a hot spare would have to be as big or bigger than any of the other drives, this would represent potential storage space (and money) not being used.
b) because "shares" currently represent disk slots, it's a bit "messy", in terms of user configuration, to implement. We have plans to upgrade to a more flexible scheme for shares which would accomodate a hot spare easier.

So what happens if the flash drive gets messed up? Bye Bye data?

Worst case, if the system can't boot and you absolutely need access to your data you have a couple options:
a) you could physically move a drive(s) to another system
b) you could email us and we'll FedEx a replacement Flash to you under warranty.

Note that a failed Flash can not cause user data loss from the hard drives.

Xezoid

09-28-05, 05:04 AM

I'm verrrry interested on this one. I prefer the software option though. Since I already have the hardware.

jimwhite

09-28-05, 05:44 AM

"even Raid 3,4, and 5 cannot recover from 2 failed drives...."

"If you have configured hot spares they sure can."

naturaly I meant simultaneous failures.... I'm sure this system could also be configured for hot spares.... but even hot spares cannot cover all bases.... if a drive fails and the hot spare kicks in and starts an automatic rebuild, any drive failures during the rebuild are fatal...

:cool:

GodAres

09-28-05, 06:42 AM

Not sure if anyone has seen this yet, but thought I'd share. This device is very similar, but good luck finding the DS-1200F (Firewire + RAID backplane), all I could find was the DS-1200U (USB w/o RAID).

Norco DS-1200 (http://www.norcotek.com/DS-1200.php)

The non-raid USB version can be found at NewEgg for a competitive price, but there's no redundancy and all the drives appear as individual USB drives, which is not what I'm looking for. I would personally love a chassis with 16 x 3.5" bays using a SATA backplane that appears as a single drive (preferably RAID 5) via USB 2.0, FireWire, or gigabit ethernet. All that I've found with those kind of features are enterprise class devices that require a second mortgage. But then again, I haven't looked into how partitions larger than 2TB can be handled. Oh well, I'm sure I'm dreaming for a device like that on the cheap anyway. A hardware RAID card that can support 16 drives is pricey in itself!

EricD

09-28-05, 10:38 AM

naturaly I meant simultaneous failures.... I'm sure this system could also be configured for hot spares.... but even hot spares cannot cover all bases.... if a drive fails and the hot spare kicks in and starts an automatic rebuild, any drive failures during the rebuild are fatal...

From what I read about this system I dont see how. It has one parity drive. There is no real RAID in the system, and its software based as well. If you change your OS, its gone. If something happens to your OS, its gone. If your parity drive goes, you better not have one of your other drives go too as it would be gone. It is such a watered down version to "save money" that its not even worth having this "unraid".

Sounds like the system is nothing more than a basic linux box with the lowest form of software raid which provides minimal, if any, data protection.

J. L.

09-28-05, 11:27 AM

From what I read about this system I dont see how. It has one parity drive. There is no real RAID in the system, and its software based as well. If you change your OS, its gone. If something happens to your OS, its gone. If your parity drive goes, you better not have one of your other drives go too as it would be gone. It is such a watered down version to "save money" that its not even worth having this "unraid".

Sounds like the system is nothing more than a basic linux box with the lowest form of software raid which provides minimal, if any, data protection.

From what I've read you are mistaken.

Worst case scenario, if you lose the parity drive AND another drive you can still mount any of the working drives using the NT driver available at the link on the Limetech web-site, or mount any of the working drives on any other linux box and all the data on those drives would be accessible. Worst case scenario, I need to re-rip one disks worth of DVDs once I replace the failed drives... and that is if I have two hard disks fail at the same time. In this situation the unRaid is better than almost any other "Raid 4 or 5" implementation since the data is accessible from any drive that has not failed.

On any standard "striped" raid (hardware or software) if you lose the parity drive and ANY other drive, you lose ALL the data on ALL drives as none have full copies of your data. In that way, the unRaid has a major advantage.

In both cases, Raid or unRaid, if you lose a single drive, all the data is available. The goal is to replace the failing hardware as soon as possible after the failure is detected to reduce the possibility of multiple drives failing concurrently. Now, don't get me wrong, all drives will eventually crash, it is only a matter of time and MTBF. That is why I think the failure notification on the LAN is a top priority for Tom to add to his distribution.

Yes, it is software raid, but today's processors can keep up with the task so I don't see that as an issue.

I love the idea that the array can be easily expanded as needed and from what I read, it can even be used while it is being expanded. I love the idea that drives will spin down when not accessed in a while. For home media server use, this is ideal as it minimizes power consumption. For my own home, at best two or three drives will be accessed at a time, regardless of the number in the server. This too is different than standard "striped" raid where all disks must be spinning to access any file.

Now, I agree, a hot spare would improve reliability, but as long as a notification of a failed disk is attended to in a reasonable timeframe, I would rather use that same disk for media storage rather than use it as a hot spare. I mean, most of what I will store will be images of the DVDs in my collection and sitting on my shelf. The remainder will be pictures and documents I do not want to lose as a result of a disk failure. Again, since I can mix and match drive sizes (something you cannot do in most Raid arrays), I can use an existing smaller drive for those files.

I'll be placing my order soon... I think it is perfect for how I'll use it and provide the flexibility I'll need.

Joe L.

madpoet

09-28-05, 11:51 AM

One point I'm not sure you're seeing Eric is that these drives are not treated as a single volume. They are individual drives, which is why you would be able to take one out and read it in another machine.

limetech

09-28-05, 12:27 PM

From what I read about this system I dont see how. It has one parity drive. There is no real RAID in the system, and its software based as well.

Our UnRaid system of course is based on real RAID, we simply don't stripe user data across all the hard drives. This provides much greater immunity to catastorphic data loss vs. other Raid systems because you can always access each drive individually.

Regarding being software based: all RAID systems are software based. It's just a question of where the software resides. The only true hardware based function for RAID is an XOR engine, but with today's memory and bus speeds, this is completely unnecessary for all but the most demanding high bandwidth applications.

If you change your OS, its gone. If something happens to your OS, its gone.
Unless I'm misunderstanding what you mean, this is simply untrue.

If your parity drive goes, you better not have one of your other drives go too as it would be gone.
This of course is true of any other Raid-3, -4, or -5 system.

It is such a watered down version to "save money" that its not even worth having this "unraid". Sounds like the system is nothing more than a basic linux box with the lowest form of software raid which provides minimal, if any, data protection.

Our idea was to design a platform for inexpensively connecting as many hard drives as possible to a network to provide protected storage for digital media. As mentioned before, the UnRaid product is not ideal for all applications (and we make no claim that it is). But for anyone who wants to build a large on-line library of typical digital media (pictures, music, and video), we feel this product (and platform) is ideal.

EricD

09-28-05, 02:14 PM

I'd mirror your parity disk if you could, or make it an option.

J. L.

09-29-05, 12:27 AM

I'd mirror your parity disk if you could, or make it an option. Eric,

I'm not sure why you feel mirroring the parity disk would help with reliability, other than it is more heavily accessed than any other drive.

If the parity drive fails, you can still read and write to any of the storage drives, you simply will not be protected from a second disk's concurrent failure. The goal is to replace any failed drive as soon as possible and reduce the window of time where two drives could fail at the same time.

Now...I could still have a disk controller card fail and take two or four drives out of service at a single blow. In that situation I'd much rather replace a $30 disk controller card than a $300 raid controller card and get everything back up running, the data on the disks is not lost. Any motherboard or memory problem could keep you from getting to your data, but again the data itself is not lost...

From what I understand, the parity drive will get more use than any other drive in the server. I expect I'll swap out the parity drive in a year or two when even larger drives become more affordable. The larger drive will become the new parity drive and the old parity drive will become media storage. Then it will spend most of its time spun down. That should even out the usage somewhat.

So... why do you feel I should mirror the parity drive? That would NOT protect me from two other data drives from failing concurrently... nothing would protect me from that other than full mirroring of all drives. (very expensive)

Joe L.
PS
I placed my order with Limetech this evening... should be fun in the next few weeks...

szsori

09-29-05, 02:42 AM

I think it's also important to note that adding enough drives to have a hot spare for each would push the price of this unit up to enterprise storage units, which are already available through other companies. I don't know about anyone else, but when I've had a RAID-5 drive go bad at work, I immediately cut off user access until I swap the drive out. I figure the fewer reads and writes in that 5 minute period, the less of a chance I'd be forced to recover from tape.

I'd treat this device in the exact same way. If you want to make it as effective as possible, keep 2 spare drives on hand to swap out if you ever have problems. I'd personally shut the whole thing down until a replacement drive was in there. Perhaps that could be an option in the software? Once a notice is sent out regarding a failed drive, the whole thing could power down. That would reduce the chance that another drive would go bad before you put in a spare.

I'm waiting on the announcement coming up in a few days. Hope the software version doesn't cost more than the wife is willing to let me spend. :) Lots of spare hardware on hand, so a software version is ideal for me.

limetech

09-29-05, 03:08 AM

... are you planning an 2.6 kernel version? If so, when. Would I be able to run all the current services I have now?

Yes we'll be migrating to 2.6 as soon as we've complete the SATA version of the product; though a 2.6 based product probably won't be available until Dec/Jan time frame.

Regarding running other services on the server: maybe. One problem is that we don't designate any swap space. So if a process consumes too much memory the kernel starts killing other processes, eventually resulting in a crash. People experienced with linux could sparingly install other services in the current product, but probably not the entire list you enumerated.

limetech

09-29-05, 03:21 AM

...Perhaps that could be an option in the software? Once a notice is sent out regarding a failed drive, the whole thing could power down. That would reduce the chance that another drive would go bad before you put in a spare.

That's a good idea.

MrSensitive

09-29-05, 05:47 AM

That's a good idea.

I'm not to sure about that..
keeping in mind that drives usually fail when starting..

J. L.

09-29-05, 07:45 AM

I don't know about anyone else, but when I've had a RAID-5 drive go bad at work, I immediately cut off user access until I swap the drive out. I figure the fewer reads and writes in that 5 minute period, the less of a chance I'd be forced to recover from tape.

I'd treat this device in the exact same way. If you want to make it as effective as possible, keep 2 spare drives on hand to swap out if you ever have problems. I'd personally shut the whole thing down until a replacement drive was in there. Perhaps that could be an option in the software? Once a notice is sent out regarding a failed drive, the whole thing could power down. That would reduce the chance that another drive would go bad before you put in a spare.
szsori,
I think that in a home media server role, especially with drives that automatically spin down when not in use, we have far more control than in a corporate environment. as you said, we can unplug the server if we're that worried and play DVDs the old-fashioned way.

Since the "cold-spare" drives would have to be as big as your two largest drives and since they could be kept "hot" by using them in the array (assuming it is not maxed out) I'd personally rather store another 100 movies online.

It is impossible for me to guarantee I can replace a failed drive within "5 minutes" (there are times when I spend 5,10, or even 15 minutes away from computers on my LAN :D)

keeping in mind that drives usually fail when starting..MrSensitive,
I don't have enough experience with todays drives to know how they fail most frequently. I've had about 6 or 8 disk failures over the past 15 years in computers in my home. Most ended up still spinning, but unable to get to some of the data as a disk head crashed. I can only remember one that failed when it stopped spinning the disk.

For critical data, priceless recordings, documents, and pictures that would be impossible to replace I make backup copies of the files to writable CDs/DVDs. I also make full backups of my machines "C:" drives and save those backups to other shared drives on my network. The unRaid (or any form of Raid) is not a substitute for these backups.

Joe L.

J. L.

09-29-05, 08:02 AM

EricD

09-29-05, 12:54 PM

So... why do you feel I should mirror the parity drive? That would NOT protect me from two other data drives from failing concurrently...

The parity drive is used the most and is the most important to other drives (containers) if they fail. Are you standing at your raid array 24/7 to see if a drive fails so you can immediately replace it? Since there is no hot spare feature in this device I suggested mirroring the most important drive.

I think it's also important to note that adding enough drives to have a hot spare for each would push the price of this unit up to enterprise storage units

For each what, drive? You have one hot spare for all drives, so its one lossed bay is all.

madpoet

09-29-05, 02:11 PM

The parity drive is no more or less important than any other drive in the system Eric. It just has the most activity.

szsori

09-29-05, 02:34 PM

I'm not to sure about that..
keeping in mind that drives usually fail when starting..

Everything I've read this system doesn't have hot-swap capability anyway. That means you have to shut it down to put in a new drive. Why not have the option to have it shut itself down if it notices any problems? The chance that another drive goes bad before you're able to put in a new drive is pretty slim, but in my situation I can go without my HTPC for a day or two to ensure I'm not going to lose a whole drive of data.

Besides, while starting a drive spinning puts some strain on the drive (hence why many people want their drives spinning constantly), a drive is just as likely to fail when you're not restarting the system. Most drive failures are caused by high heat over time and brownouts, which are gradual failures. Most sudden failures are caused by power outages, head failures, physical shock, and power surges. By shutting down the system, you're reducing heat, brownout effects, sudden power outages, head failure, and power surges (to some extent). Just don't kick the tower and you should be safer shutting it down. :)

Eric: You're right, one hot spare would be all that's needed, as long as it can automatically take over for any drive (not just the parity drive). I think you're misunderstanding how a parity set really works, though. The parity drive is no more important than any other drive, since the entire set of drives is required to rebuild a single drive that goes bad. The parity data is calculated by comparing the data on ALL of the drives.

In fact, the parity drive is actually LESS important than the data drives in this type of setup. If you lose one drive, you can recover no matter what. If you lose 2 drives, one being the parity drive and another being a data drive, you lose one drive worth of information. If you lose 2 drives, both being data drives, you lose two drives worth of information.

EricD

09-29-05, 02:54 PM

The parity drive is no more or less important than any other drive in the system Eric. It just has the most activity

So lets say your parity drive goes down. Then if ANY other drive goes down, you loose its information. It is more important.

If, instead, one normal (container) drive goes down. Then if any other drive BUT the parity drive goes down, you should be fine.

Im just trying to give ideas if TWO drives fail is all. Granted, I come from standard raid 0/1/5/10 experience. Im still shaky how this 'unraid' writes the info. The webpage says no data is striped across the drives, like a typical raid4 would be. So if one standard drive fails, it looks at the parity drive for the rebuild info but where else for the data?

His product is nice for home use, but a little expensive for some. Other issues would definitely have to be addressed before it goes to a business environment.

J. L.

09-29-05, 02:59 PM

The parity drive is no more or less important than any other drive in the system Eric. It just has the most activity.madpoet,
In a media server role I'm not sure the parity drive is accessed when reading a data drive. There is no need to access the parity drive when reading data (serving movies) unless the read of the data drive reports a failure. Only then is there a need to spin up the parity drive and all the other data drives to reconstruct the missing data that could not be read from the failed data drive.

So... I'll agree, when initially loading the DVD images onto the data drives the parity drive is plenty busy (equal to the drive being loaded), but when serving up movies, it might not be the drive with the most activity... in fact, if everything is working properly, it might be the one with the least.

szsori,
I can see where you are coming from, but if I'm watching a movie I'd rather be notified of the failure and be able to finish watching the movie with the unRaid array doing what it was designed to do. I'll shut the array down when it is appropriate for me...

In that respect, the option to automatically shut down should be configurable if it exists. I would not normally want it immediate. Perhaps a configurable delayed shutdown would work better. (Warning... media server disk failure... system shutdown in 2/4/6/8 hours)

Joe L.

madpoet

09-29-05, 03:24 PM

Probably true Joe, I was thinking in terms of the overall "loading" of the box.

Eric, if any 2 drives fail you lose data from my understanding of the system. Like almost every other raid solution, it's only meant to support a single drive failure. It wouldn't matter if it was the parity drive and another drive, or 2 other drives.

Thomas J. Coyle

09-29-05, 03:33 PM

By szsori:
"Most drive failures are caused by high heat over time and brownouts, which are gradual failures. Most sudden failures are caused by power outages, head failures, physical shock, and power surges. By shutting down the system, you're reducing heat, brownout effects, sudden power outages, head failure, and power surges (to some extent)."

Anyone who operates a server and does not not power that server from an automatic line regulation UPS of sufficient size and reliability, is asking for trouble in the first place.

Regards,
TCIII

jimwhite

09-29-05, 03:35 PM

"If, instead, one normal (container) drive goes down. Then if any other drive BUT the parity drive goes down, you should be fine"

you still don't get it..... if ANY 2 drives go down, you lose data.... doesn't matter wether it is parity or dasta drive.... EXCEPT in un-Raid's case, as noted above:

"If you lose 2 drives, one being the parity drive and another being a data drive, you lose one drive worth of information. If you lose 2 drives, both being data drives, you lose two drives worth of information."

:cool:

xradman

09-29-05, 03:42 PM

EricD

09-29-05, 05:38 PM

If you lose 2 drives, both being data drives, you lose two drives worth of information

Like I said, this doesn't make sense in that they said the drives arent striped in any way. Are they mirrored with a common parity drive? If so, it would all make sense. From their site:

" Unlike other RAID organizations, however, files are not striped across the data drives."

This statement, along with all the other stuff that contradicts it, is what confuses me.

jimwhite

09-29-05, 06:29 PM

Back to basics....

1) if you lose any one drive, the missing drive can be rebuilt

2) if you lose the parity drive and a data drive simultaneously, you've lost one disk worth of data, no rebuild possible...

3) if you lose 2 data drives simultaneously, you've lost 2 drives worth of data, no rebuild possible

on a classic raid 3 or 4 or 5 system,

1) same results
2) all data gone
3) all data gone

make sense?

:cool:

EricD

09-29-05, 06:57 PM

3) if you lose 2 data drives simultaneously, you've lost 2 drives worth of data, no rebuild possible

Why? Yes, true in standard Raid 4/5 etc. This is because data is striped in standard raid 4, that its based on. However, straight from their site:

" Unlike other RAID organizations, however, files are not striped across the data drives."

I'm trying to find out how it actually writes the data to the disks without striping.

J. L.

09-29-05, 09:35 PM

if you lose 2 data drives ...snip... no rebuild possible.
...snip... This is because data is striped in standard raid 4, that its based on. Actually, the inability to rebuild has nothing to do with striping... I'll attempt to explain.

As jimwhite already said, losing two drives in a standard raid 4/5 array results in a loss of the data on ALL of the drives in the raid array, not just that stored on the two specific drives that failed. So if you had 11 drives of data and one parity drive and lost any two of them then ALL of the data on ALL of them is gone.

However, straight from their site:

" Unlike other RAID organizations, however, files are not striped across the data drives."

I'm trying to find out how it actually writes the data to the disks without striping. Striping data across drives was done when raid arrays were initially developed so reading data off drives could be done in parallel to speed up access. As Tom has pointed out, this is not a limitation today in the drives we can purchase and use in a home media server. Basically, the first block of bytes in a given file was written to the first disk in the raid array, the next block of bytes to the second, the next to the third, etc wrapping around to the first, second, third, etc. as needed until the whole file was stored.

If you had 5 data drives then one fifth of your file was stored on each drive in a way that you could issue 5 read commands simultaneously, one to each drive, and fetch the first 5 blocks of data in parallel. As soon as the first drive had read its block it could be given the command to read its next chunk of the file (the 6th block) and so on...
Disk density used to be much lower and disk access slower. The parallel reads of data was a great way to speed access to large files. Striping data across drives increases speed of access... nothing more. It has nothing to do with making the data tolerant to a drive failure.

So, unless your file was very small in Raid 4 or 5 it is spread across the physical disks and losing any physical disk would result in losing the ability to get to one fifth of any file stored in the array. In effect, a loss of any drive results in a loss of every file since none of the files are completely available. All disks must be constantly spinning as to get to any file since every disk must be read to get all the pieces "stored/striped" accross them.

If instead of "striping data across drives" we stored whole files on specific drives then the loss of one specific drive would result in only the loss of the files on that drive. The files on the other drives are still completely available. This is exactly how most disks are configured in our desktop PCs. To read any given file, only the single drive it is stored on must be spinning. The other drives can be spun down for reduced power consumption if desired in periods of inactivity.

Notice I said nothing about the parity drive and how we can add one to a Raid array to increase reliability. Its job is so we can store a set of values that allow us to recreate any single missing drive. To make things simple in the following example, assume 3 data drives, plus one more used for parity. Now we can either use ODD or EVEN parity in our design... lets use ODD. To keep things even easier, this fictional raid array has a storage capacity of one bit. A single 1 or zero can be stored in each drive.

To calculate parity we simply add up the number of "1's" and if the number is even we set the bit on the parity drive to make the total number of bits set to "1" ODD.

So
Data (one bit from each drive) = 100 Parity = 0 ( total number of 1's, including parity bit = 1)
Data (one bit from each drive) = 000 Parity = 1 ( total number of 1's, including parity bit = 1)
Data (one bit from each drive) = 010 Parity = 0 ( total number of 1's, including parity bit = 1)
Data (one bit from each drive) = 001 Parity = 0 ( total number of 1's, including parity bit = 1)
Data (one bit from each drive) = 110 Parity = 1 ( total number of 1's, including parity bit = 3)
Data (one bit from each drive) = 101 Parity = 1 ( total number of 1's, including parity bit = 3)
Data (one bit from each drive) = 011 Parity = 1 ( total number of 1's, including parity bit = 3)
Data (one bit from each drive) = 111 Parity = 0 ( total number of 1's, including parity bit = 3)

If we lose any single drive we can deduce the missing value since it must result in an ODD number of 1's.

To simulate a failed drive I'll simply show a ? in the place where its bit would be...
Data (one bit from each drive) = 1?0 Parity = 0 ( to get an ODD number of 1's, the missing value would be 0)
Data (one bit from each drive) = 0?0 Parity = 1 ( to get an ODD number of 1's, the missing value would be 0)
Data (one bit from each drive) = 0?0 Parity = 0 ( to get an ODD number of 1's, the missing value would be 1)
Data (one bit from each drive) = 0?1 Parity = 0 ( to get an ODD number of 1's, the missing value would be 0)
Data (one bit from each drive) = 1?0 Parity = 1 ( to get an ODD number of 1's, the missing value would be 1)
Data (one bit from each drive) = 1?1 Parity = 1 ( to get an ODD number of 1's, the missing value would be 0)
Data (one bit from each drive) = 0?1 Parity = 1 ( to get an ODD number of 1's, the missing value would be 1)
Data (one bit from each drive) = 1?1 Parity = 0 ( to get an ODD number of 1's, the missing value would be 1)

Parity can be calculated across a set of drives regardless of what is stored on them... if the block on a given drive is not allocated, all its bits are zero. It doesn't even matter how many disks are in the raid array... we only need one parity drive and set its bit to get an ODD number of bits set to a "1" across the entire set of disks.

Now... losing two drives would result in the inability to deduce the two missing values...
Data = 1?? Parity = 0 ( to get an ODD number of 1's, the missing value could be either "00" or "11" both would result in an odd number of 1's, but one of the two is wrong... we cannot recreate the missing information from the failed drive)

Data = 0?? Parity = 0 ( to get an ODD number of 1's, the missing value could be either "01" or "10" both would result in an odd number of 1's, but one of the two is wrong... we cannot recreate the missing information from the failed drive)

Now... obviously, each disk drive each holds more than one bit, but the process is exactly the same for every bit they do store. The equivalent bits on the other drives are used to calculate and store a single parity bit (on the parity drive) for every bit position on a data drives. That is why the parity drive must be as big or bigger than all the data drives... so it will have an equivalent bit position to set as an indication of parity.

If we lose the parity drive we can recreate it by reading the bits from the data drives and setting the equivalent bit to 1 or 0 as needed to result in either ODD or EVEN parity. (As I said, any given implementation could use either... if EVEN parity is used the parity bit is set to result in an EVEN number of bits set to a "1")

So... striping results in faster access to data... and in the absence of a parity drive results in ALL your data being lost if ANY data drive fails. With a parity drive, it results in faster data access AND tolerance of any single drive failure.

Not striping results in slower data access... but still way fast enough with today's PCI bus, disk controllers, and disks for feeding multiple streams of video in a home media server. In the absence of a parity drive, loss of a single drive results in the loss of that single drive's files. With a parity drive you get tolerance of any single drive failure. This is how Tom's "unRaid" is configured.

With parity and striping (Raid 4 or 5), loss of any two or more drives results in loss of all data on all drives.

With parity and not striping (unRaid) loss of any two or more drives results in loss of data on the lost drives only. Data on the other drives is still available.

Joe L.

limetech

09-30-05, 02:02 AM

Limetech,

Assuming I do keep a "pseudo-hot-spare" drive in one of the slots in the unRaid server, and assuming I store stuff on it that I don't mind overwriting if I need to press the drive into service as a new "parity drive," how would I go about reusing a data drive as a replacement "parity drive?"

Is it as simple as putting it in the top slot in the unRaid server? (did not see the answer to this in your FAQ)

Joe L.

Yes. In the case where the parity drive has failed, you can move one of the other drives into the parity slot, go to the Tools menu and "Initialize array configuration", then Start the array - parity will get generated using data from the remaining data drives.

limetech

09-30-05, 02:08 AM

Is it possible to run something like VNC on the server so that you have Mon/Key board/Mouse control from a remote computer? I can see this as being useful for accessing local server functions and also for accessing remote drives and accessories one might attach to the Firewire or USB ports.

You can use telnet to remotely control and monitor the system, though the browser-based management utility is the best way to manage the array.

limetech

09-30-05, 02:13 AM

limetech

09-30-05, 02:16 AM

Actually, the inability to rebuild has nothing to do with striping... I'll attempt to explain.

*snip*

Joe L.

Exactly right. BTW, we use even parity.

limetech

09-30-05, 02:48 AM

...I'm trying to find out how it actually writes the data to the disks without striping.

Here's perhaps an easier way to think of it. Suppose you have Just a Bunch Of Disks. Each disk is accessed independently of all the other disks and each has it's own file system, and perhaps even different file system types.

Now add another disk and call it parity. Then if your OS had the ability you could tell it to generate parity by reading block 0 from each of your data disks, XOR them together, and write that to block 0 of the parity disk; then repeat for all the other blocks. When you've reached the end of a data disk, start supplying "zeros" for it in the parity calculation. Now you have the parity information Joe explained previously.

Thereafter, any time you write a block on one of your data disks, the OS would do this:
a) first read the curent contents of the block you're about to write (call this "old data")
b) read the corresponding parity disk block (call this "old parity")
c) write your "new data" to the data disk block
d) XOR together your "old data", "new data", and "old parity" to generate "new parity" which is then written to the parity block.

This is exactly the algorithm performed by single-block Raid-4 and Raid-5 writes and accounts for the so-called "write penalty" in these systems.

As an aside: true hardware raid cards can speed this up in a couple of ways. First, they can perform the XOR calculation "on-the-fly", which means as data is moving between a disk and memory, the hardware can XOR it with a block already in memory. Second, this "memory" can be on-board which eliminates having to use your system memory and PCI bus for doing this calculation. Note however that you still have two disk reads and two disk writes which are orders of magnitude slower than the memory operations.

EricD

09-30-05, 11:29 AM

If instead of "striping data across drives" we stored whole files on specific drives then the loss of one specific drive would result in only the loss of the files on that drive.

I understand normal raid just fine. Its rebuilding from this unraid where I am confused (yes, still). You say its based on raid4. I understand raid4 in that the data is spread accross the drives and the parity is stored on a separate drive. If a drive fails, it has enough information from the other drives and the parity info to rebuild. With the unraid, you say ALL of the file data is stored on ONE drive. From where do you get the data back from? From only the parity information? Is it all in the algorithm with the parity info that you can restore all the data? Thats not standard parity from what I remember.

yakking

09-30-05, 12:42 PM

I understand normal raid just fine. Its rebuilding from this unraid where I am confused (yes, still). You say its based on raid4. I understand raid4 in that the data is spread accross the drives and the parity is stored on a separate drive. If a drive fails, it has enough information from the other drives and the parity info to rebuild. With the unraid, you say ALL of the file data is stored on ONE drive. From where do you get the data back from? From only the parity information? Is it all in the algorithm with the parity info that you can restore all the data? Thats not standard parity from what I remember.

This is what I understand of parity:

Block 0
Drive 1: 0
Drive 2: 1
Drive 3: 0
Drive 4: 1
Drive 5: 1
Drive 6: 0
Then you add them up (basically)
Parity Drive: 1
(0+1+0+1+1+0 = 11, I guess we just take the last digit?)

If you lose a drive, say drive 3, then you have this data:
Block 0
Drive 1: 0
Drive 2: 1
Drive 3: x
Drive 4: 1
Drive 5: 1
Drive 6: 0
Parity Drive: 1

This time you use some fancy subtraction and you find that Drive 3 Block 0 is 0.
(0+1+x+1+1+0 = 11+x. Parity is 1 and take just the last digit, so 1+x=1. Therefore x = 0.)
Then you rebuild and work on the other blocks.

Am I right?

*edit*
I forgot to mention that this shouldn't matter whether drives 1-6 are striped, because the physical data is still there. There's a Block 0 on all the drives, and so forth, up to the highest block of the largest drive. That's why the parity drive needs to be as large as the largest drive. The only difference between striped and unraid is that striped as all data across all the drives. Whereas unraid keeps the individual drives separate with their own data. Parity would still check through all the drives.

J. L.

09-30-05, 12:54 PM

Is it all in the algorithm with the parity info that you can restore all the data?EridD,
Yes... it is the algorithm.

I'll try a new example... Lets say I set up my new UnRaid server with 5 disks for data and 1 for parity.

For arguments sake:
the very first file located on the first disk is a text file joe.txt
the very first file located on the second disk is a picture of my cat
there are no files on the third disk... it is currently empty
the very first file located on the fourth disk is a MS-Word document
the very first file located on the fifth disk is a PDF document

Now... I also have a parity disk and I need to store data on it so if I lose any single drive I can still rebuild its data.

So I start by reading the first bit of each file on each disk.

the very first bit in the text file joe.txt on the first disk is a "1"
the very first bit in the picture of my cat on the second disk is a "0"
for the third disk the spot on the disk where the first file would be is a "0"
the very first bit in the MS-Word document on the fourth disk is a "1"
the very first bit in the PDF document on the fifth disk is a "1"

Now... Tom has said he stores EVEN parity... so far we have an ODD number of ones in this bit position across all the drives. Since we store EVEN parity we need to store a "1" on the parity drive for this bit position. (that will give us a total of four "1s", an EVEN number, in this bit position on the drives. ) Remember, we always set the parity bit so we have an even number of '1's for a bit position. If the second bit position on ALL the data drives in my array had a total of 2 of the bits set to a "1" the parity bit would be set to '0' as we already had an EVEN number of '1's in that bit position.

Now... lets say one of the drives fails and the unRaid software detects the failed drive... since we have a parity drive we can figure out what would have been on the failed drive if we could have read it... To do this we need to read the first bit from all the drives still available.

we read the very first bit on the first disk as a "1" (joe.txt)
we read the very first bit on the second disk is a "0" (picture of cat)
we read the very first bit off the empty third disk is a "0"
we CANNOT READ the fourth disk as it has failed - this is the bit we need to recreate
we read the very first bit on the fifth disk is a "1" (PDF document)
we read the very first bit in parity drive and it is a "1"

Now... we know the missing bit on the fourth disk could be either a "0" or a "1" and also know that when we originally calculated and stored the bit on the parity drive we had an EVEN number of "1s" in that bit position on the drives.
So... with that knowledge we can figure out what the missing bit would have been if we could have read it.

Since in all the bits we can still read (data bits in the working drives and the bit in the parity drive) we have an odd number of "1s", the missing bit we cannot read because of the failed drive must have also been a "1" to make it an even number of "1s."

The bit on the failed drive can't be a "0" since we would then only have three bits set to "1" in that bit position, (an odd number) and since know we originally calculated and stored EVEN parity... the missing bit had to be a "1", for a total of 4 bits set to "1" (an EVEN number)

So... the algorithm does let us recreate any single missing bit by using the data on the parity drive in combination with all the other data drives that can still be read.
Use the same algorithm for each bit in turn on the entire set of drives and you can recreate the entire set of bits that existed in the failed drive.

Joe L.

EricD

09-30-05, 01:18 PM

Thanks Joe

szsori

09-30-05, 02:03 PM

Just a quick aside. If you want to create parity information for folders or files on your drive, but on a much smaller scale, check out the PAR2 program (Google it). It allows you to create parity files for files and folders. You can then save the parity files to CD or DVD backup instead of having to store a backup of all of the files themselves. When I burn avi's to DVD (3 per), I generate parity information and store it on the drive as well (I encode avi's to make sure I leave 10% of a DVD open). This way I can recover from most burn errors and disk rot. It's saved a couple of files already.

I tried doing the same thing that the Unraid software does, but with the par2 command line program once a night. It takes WAY too long (days of computation). That's why I'm waiting for the software-only announcement from limetech (tomorrow, right?). :)

musicmann

09-30-05, 02:15 PM

Why? Yes, true in standard Raid 4/5 etc. This is because data is striped in standard raid 4, that its based on. However, straight from their site:

" Unlike other RAID organizations, however, files are not striped across the data drives."

I'm trying to find out how it actually writes the data to the disks without striping.

This is the way I understand it. The drives appear as separate shares on your network. So a fully maxed system (11 data + 1 parity drive) would present 11 shares to the network. You choose a SHARE to write to, which is equivalent to choosing one disk to write to. Plus, when a (data) disk is written to, parity must be calculated and written to the parity disk.

True, this is not the most efficient way of allocating storage (thinking about as you near capacity on each drive), but it is a reasonable tradeoff given the fact that losing multiple drives does not mean losing ALL of the data.

Thomas J. Coyle

09-30-05, 04:51 PM

limetech,

Do you still plan to make your announcement today concerning s/w operating system only sales? I am really chomping at the bit to get it checked out on my CM Stacker DVD server before I have to fall back on Windows XP Pro!
Regards,
TCIII

Thomas J. Coyle

09-30-05, 07:21 PM

limetech

09-30-05, 07:32 PM

This is the way I understand it. The drives appear as separate shares on your network. So a fully maxed system (11 data + 1 parity drive) would present 11 shares to the network. You choose a SHARE to write to, which is equivalent to choosing one disk to write to. Plus, when a (data) disk is written to, parity must be calculated and written to the parity disk.

Exactly right.

True, this is not the most efficient way of allocating storage (thinking about as you near capacity on each drive), but it is a reasonable tradeoff given the fact that losing multiple drives does not mean losing ALL of the data.

Again, exactly right.

We are considering offering an optional configuration setting that would let you view all the storage as a single large volume. We won't be able to do this until Dec/Jan time frame. Here's the tradeoff if we do this:
- if you lose two drives, you may lose ALL your data :mad:
- you won't be able to replace a small drive with a larger one :(
Personally, I would never use this option, but I can see how it could be definitely attractive for some applications.

limetech

09-30-05, 07:33 PM

limetech

09-30-05, 07:43 PM

limetech,

I have looked at the hard drive installation instructions on your web site and the mobile hard drive docks do not seem to have fans. I have been using mobile docks with fans and even with fans, the drives will approach 35 degrees C on a warm day (around 75-80 degrees F) when they have been on for a long time. I have all of the cooling fans installed in my CM Stacker so I am not lacking for air flow through the case itself. Do you feel that it is not necessary to have the additional cooling of fans in the mobile docks?
Regards,
TCIII

It's not obvious from the photos, but the trays DO have fans. But here's what we've found after testing probably 20 differennt mobile racks. In many mobile racks the fans don't make a huge difference. Using the CM Stacker you could probably remove all the fans and tray lids and it would be just fine.

Thomas J. Coyle

09-30-05, 07:47 PM

limetech,

Thanks for the update. Much appreciated.
Regards,
TCIII

Birdman71

09-30-05, 09:12 PM

Are we there yet?

Are we there yet?

Are we there yet?

Thomas J. Coyle

10-01-05, 09:17 AM

limetech,

Do you still intend to support the Marvell 88E8001 that is used on the Asus AMD motherboards (A7N8X-E Deluxe) or are you only going to support the Marvell 88E001 (By the way, Marvell does not list this part on its web site)? I will definitely buy the s/w operating system if you can provide support for the more common Marvell 88E8001, otherwise I will have to stick with Windows XP Pro which I already have available for use on my second DVD server.
Regards,
TCIII

limetech

10-01-05, 03:42 PM

Are we there yet?
Almost....

Are we there yet?
Just one more minute.....

Are we there yet?
Yes, finally! Still more info to post, but the product list is ready.

limetech

10-01-05, 03:44 PM

J. L.

10-01-05, 05:54 PM

Almost....

Just one more minute.....

Yes, finally! Still more info to post, but the product list is ready.Tom,

When I clicked on the link in your signature I got the old page.

I had to force my browser to clear its cache (control refresh) in order to see the new product page.

But... the product offerings are there now... and look good.

Joe L.

szsori

10-01-05, 09:01 PM

The 2 new product offerings look great! A few questions:
- What is the minimum amount of RAM needed? I'd probably start with the minimum and add more around Christmas, so I guess minimum and recommended amounts.

- What's the minimum speed processor needed? I'd probably go with one of the motherboards you recommend, so the minimum is around a 2Ghz anyway. Is that sufficient?

- If I don't put the two controller cards in right away, will it work using the 2 onboard connections? When I would add the controller cards in the future, will it need any special configuration or will it just autodetect and work?

- How do you determine the drive numbering (ie.. which is the parity drive) in a setup with no controller cards? When you add on controller cards will those just become 5-12?

- How much would you sell JUST a Coolermaster Stacker case, in the event I'd upgrade to it in the future?

Thanks!

Thomas J. Coyle

10-01-05, 09:40 PM

szsori,

If limetech does not sell the CM Stacker by itself, it can be purchased from either mwave.com or newegg.com. The last CM Stacker that I bought came from mwave.com as they had the best price and shipping cost.
Regards,
TCIII

Thomas J. Coyle

10-01-05, 09:47 PM

Limetech,

Thanks for the Marvell part number correction. I will be placing an order tonight for the s/w package. I am presently using an Asus A7N8X-E Deluxe that uses the Marvell 88E8001 controller so hopefully it will work with your software.
Regards,
TCIII

limetech

10-02-05, 04:19 AM

The 2 new product offerings look great! A few questions:
- What is the minimum amount of RAM needed? I'd probably start with the minimum and add more around Christmas, so I guess minimum and recommended amounts.
Minimum and recommended is 512MB, "known-to-work" maximum is 1GB, absolute maximum is whatever your card supports (typically 4GB). I say "known-to-work" because we've actually tried it and had no problems. Beyond 1GB linux uses a trick called "highmem support". Our software should work with more than 1G of memory, but this has not been thoroughly tested.

BTW, for m/b's which feature "dual channel" memory architecture, you get the best performance by installing 2 matched memory sticks. "Matched" means same size and type. So for 512MB you would want 2 sticks of 256MB.

- What's the minimum speed processor needed? I'd probably go with one of the motherboards you recommend, so the minimum is around a 2Ghz anyway. Is that sufficient?
Yes.

- If I don't put the two controller cards in right away, will it work using the 2 onboard connections?
Yes.

When I would add the controller cards in the future, will it need any special configuration or will it just autodetect and work?
It autodetects.

- How do you determine the drive numbering (ie.. which is the parity drive) in a setup with no controller cards? When you add on controller cards will those just become 5-12?
There is an editable configuration file on the Flash which specifies the relationship between controller ports and disk slots. We'll be posting the documentation which ships with the Flash card on our web site in the next couple of days.

- How much would you sell JUST a Coolermaster Stacker case, in the event I'd upgrade to it in the future?
We won't be selling this as a separate product. This case is available from a number of online retailers, e.g., newegg. We would not be able to beat their prices.

madpoet

10-02-05, 08:16 AM

Hey guys, I admit I was an early skeptic but I'm impressed. I do have some questions though:

1) If you're only using 2 controller cards and the 2 onboard IDE controllers, then how does the flashdisk controller attach? Doesn't it have to be seen as an IDE? Or can you boot straight off the flash somehow?

2) On the controller cards themselves... I notice from the barebones kit you're using Promise Ultra100 TX2 controllers. I have several of the Maxotr ATA133 controller cards which are based on the Promise PDC20269 chip. Do you require that EXACT Promise controller, or will others work? I ask because the faster transfer rate would help, and of course because I have these cards already ;). Is there a specific functionality of the TX2 controller that is important here?

Thanks,
MP

Birdman71

10-02-05, 09:49 AM

Of course it's a little early with you just releasing the product, but.........

Powerbuy?

Birdman

John Spicer

10-02-05, 09:58 AM

I may have missed this but which SATA controllers are supported? I already have the Intel D865GLCLK with a 2 ghz celeron and memory sat around, but most of my current drives are sata.

Will you ship to the UK?

Cheers

J. L.

10-02-05, 10:09 AM

1) If you're only using 2 controller cards and the 2 onboard IDE controllers, then how does the flashdisk controller attach?
Doesn't it have to be seen as an IDE? Or can you boot straight off the flash somehow?

From Tom's web-site...
The flashdisk controller is a USB device and plugs into the back of the server.
The bios on the motherboard he uses allows you to boot directly off of the USB flash drive if it has an appropriate file system and boot image to boot from.
You can see the USB flash drive on the back of the server at this link to Tom's web-site (http://lime-technology.com/Images/flash_small.jpg)

madpoet

10-02-05, 10:16 AM

bton

10-02-05, 11:00 AM

Tom,
I saw you use Promise Ultra100 TX2. Will it work with Ultra133 TX2. I already have this card.

Geordon

10-02-05, 06:08 PM

I have a Promise Ultra100 (not TX2). Will it work with this card?

limetech

10-03-05, 03:12 AM

Hey guys, I admit I was an early skeptic but I'm impressed. I do have some questions though:

1) If you're only using 2 controller cards and the 2 onboard IDE controllers, then how does the flashdisk controller attach?
The Flash drive is just a normal USB2.0 device which may be plugged directly into any one of the USB connectors on the I/O panel of the motherboard.

Doesn't it have to be seen as an IDE? Or can you boot straight off the flash somehow?
Strangely enough, linux treats it as a SCSI device. The motherboard has an option to boot from USB Flash drive.

2) On the controller cards themselves... I notice from the barebones kit you're using Promise Ultra100 TX2 controllers. I have several of the Maxotr ATA133 controller cards which are based on the Promise PDC20269 chip. Do you require that EXACT Promise controller, or will others work? I ask because the faster transfer rate would help, and of course because I have these cards already ;). Is there a specific functionality of the TX2 controller that is important here?
The Maxtor ATA133 cards do work; however, they are NOT as fast as the Promise-branded cards. I don't know why this is true and we spent a lot of time scouring code looking for a possible bug, only to discover this. I remember being so disgusted I went around our lab and collected every Maxtor card we had and pitched them in the "recycle bin".... So I don't recommend them, but I *think* you can reflash them to use the generic code provide by Promise, but I haven't tried that - no more cards left to try it on :eek:

limetech

10-03-05, 03:12 AM

Of course it's a little early with you just releasing the product, but.........

Powerbuy?

Birdman
Sure, if AVS will let us...

limetech

10-03-05, 03:16 AM

I may have missed this but which SATA controllers are supported?
I haven't listed the SATA controller card yet.

I already have the Intel D865GLCLK with a 2 ghz celeron and memory sat around, but most of my current drives are sata.
That's a good motherboard & processor. The SATA version is almost complete.

Will you ship to the UK?
We're going to shipping internationally some time this month!

limetech

10-03-05, 03:45 AM

Ahh, thanks J.L. Answers that question! :)

Thought of another one as well. Since the Parity drive will be forced to share a controller with one of the other drives, then when you are writing anything to the drive on that same channel it's going to be REALLY slow. That's the weakness of the Master/Slave IDE configuration. I realize this only affects 1 out of the remaining 11 drives, but have you seen significant issues with that?
I was wondering when someone would ask that. It turns out there is very little performance difference between disk1 (the share on the same controller as parity) and any of the other disks; but it is slower. We're re-running all our performance testing again this week & I'll include that as one of the things to quantify.

limetech

10-03-05, 03:47 AM

Tom,
I saw you use Promise Ultra100 TX2. Will it work with Ultra133 TX2. I already have this card.
Should work fine; though refer to a similar question a few posts back about the MAXTOR ultra133 tx2.

limetech

10-03-05, 03:54 AM

I have a Promise Ultra100 (not TX2). Will it work with this card?
I don't have this card, so I don't know. Probably not.

madpoet

10-03-05, 08:34 AM

I was wondering when someone would ask that. It turns out there is very little performance difference between disk1 (the share on the same controller as parity) and any of the other disks; but it is slower. We're re-running all our performance testing again this week & I'll include that as one of the things to quantify.

Thanks Tom. I know in running a RAID5 setup using Master/Slave it was a significant performance hit, but then again you're writing a stripe to all of the disks. Also, have you guys seen issues where if Master fails, Slave is also marked as failed? It happened to me a couple of times with other RAID solutions. It wasn't the end of the world because the Slave usually just had to be brought back online and was fine, but I've had several people tell me this was an issue with using Master/Slave drives in any sort of RAID.

Thomas J. Coyle

10-03-05, 09:39 AM

Tom,

Which motherboard did you use to test your operating system s/w with for the Marvell 88E8001 Gigabit LAN chip?
Regards,
TCIII

pischke

10-03-05, 04:48 PM

That's a good motherboard & processor. The SATA version is almost complete.

Will the SATA version support a mix of SATA and PATA drives? Reason I ask is that I have a number of PATA drives available now to initially construct my array, but I'd like to eventually add/replace with SATA drives.

Thanks.

-pischke

musicmann

10-03-05, 05:25 PM

...Also, have you guys seen issues where if Master fails, Slave is also marked as failed? It happened to me a couple of times with other RAID solutions. It wasn't the end of the world because the Slave usually just had to be brought back online and was fine, but I've had several people tell me this was an issue with using Master/Slave drives in any sort of RAID.

I mentioned this thread over at HTPCNews, and someone made a similar comment about a failed Master also taking out a Slave with it.

1. Is this true?
2. If so, would it be possible to load a system with a bunch of controller cards, so each drive could be a master?
3. Is this an issue with SATA drives?

mpenton

10-03-05, 10:18 PM

limetech

10-04-05, 04:16 AM

I mentioned this thread over at HTPCNews, and someone made a similar comment about a failed Master also taking out a Slave with it.

1. Is this true?
2. If so, would it be possible to load a system with a bunch of controller cards, so each drive could be a master?
3. Is this an issue with SATA drives?

There is a lot of misinformation out there about this subject. The fact is that the two drives attached to the same controller are completely independent of each other. You don't need a "master" in order to have a "slave", and the "master" does not provide any functionality whatsoever for the "slave" device. The idea that if the master fails it automatically causes the slave to fail is simply not true. The terms "master" and "slave" are merely terms used instead of say, "unit 1" and "unit 2" (actually in the ATA spec they are in fact referred to as Device 0 and Device 1). Some references:
http://www.storagereview.com/guide2000/ref/hdd/if/ide/confJumpering.html
http://www.seagate.com/support/kb/disc/ref/interface_manuals.html

What the two drives share is a single channel. The way it works is that the CPU can initiate an I/O operation on one of the drives (say Device 0, or master), and then it must wait until that operation is complete before it can initiate an I/O operation on the other drive (or the same drive). The reason why disk access is not dog slow is because the disks themselves implement read-ahead and write-behind. If you really want to slow down any system with ATA drives attached (even SATA) try turning off write behind and read ahead.

Having said that, can a device failure on a channel destroy the other device as well? The answer is a definite maybe, but extremely unlikely. What would have to happen is a catastrophic failure of the drive electronics that would somehow cause some kind of voltage surge on the cable. Note that in this kind of failure scenario the controller itself would also be toast, even if it were SATA or SCSI.

Another possiblity is a less severe electronics failure which would "hang" the channel. Sure this is possible, but again extremely unlikely (compared to typical disk failure modes). In this case the solution is to remove the failed drive. Note that this is also possible with SATA (or SCSI) as well depending on the controller chip design.

Regarding the UnRAID system - you could configure a system to use just one disk per channel if you want; but, you would only be able to install 6 disks because the Promise controller bios only supports up to 2 cards installed (2 controllers per card plus the 2 onboard controllers = 6 controllers). In my view however, this is completely unnecessary and costly. Someone might be able to make a case that the parity drive should be on a channel by itself. OK, if that's how you feel, then just don't poplulate the disk1 slot.

Where you see 1-drive-per-channel is in enterprise-class storage subsystems which must not ever fail. These are running applications such as reservation systems, banking, stock trading, etc. Even in these systems, the owners are typically so paranoid they don't even use Raid, or if they do, they mirror entire systems or have huge distributed systems.

So, for digital media files, and for a host of other applications (small enterprise backup, near-line storage, etc.), the UnRAID architecture is perfectly sound. In fact it's superior because if the two drives on a channel both go up in smoke simultaneously, you at least won't lose ALL your data.

If I still haven't convinced you, then wait for the SATA version where the illusion of drives being on separate channels is stronger :)

limetech

10-04-05, 04:32 AM

I've just purchased the software solution and already have the recommended mb & controllers. I have a couple of Northwood based processors (1.6 & 2.4 celeron) that I want to try but if they don't work I'll get the reference processor. Earlier in the the thread it was mentioned that you were looking to expand the number of mb's/processors that the software has been checked on. Do you have an automated test suite that checks for compliance in case I run the software on something other than the default system?

Those processors will work just fine. I apologize for not getting the detailed documentation on line yet - the most important components in the design are the motherboard and the PCI disk controllers:
- For m/b, you want to use Intel D865GLCLK if at all possible (or the 6 pci-slot variant).
- For PCI controllers, you want to use Promise Ultra100 or Ultra133 TX2.

As for automated test suites - as a matter of fact we do have some but they run on our hard disk based development system, not the flash. Perhaps this is something we can offer in the future.

As an aside: something that makes a good case is a simple piece of plywood. Use your m/b to mark the mounting hole locations and screw in some standoffs. Lay the disks on their sides, cable everything up and mount a fan or two and you're good to go - not a huge WAF though unless you can put it in a closet :D

madpoet

10-04-05, 08:11 AM

Thomas J. Coyle

10-04-05, 10:10 AM

Birdman71

10-04-05, 10:30 AM

musicmann

10-04-05, 10:54 AM

If I still haven't convinced you, then wait for the SATA version where the illusion of drives being on separate channels is stronger :)

I am waiting on more info re: the rebuttals/info of MadPoet and BirdMan below, but I probably will wait for the SATA version nonetheless.

1. I think you're going to win your challenge to me earlier by coming out the the SATA version before I run out of space on my current rig (600G to go!)
2. I've been trying to make the switch to SATA for a while (mostly for the cleaner cables), and with my current system maxed out, I can finally make a clean switch

mpenton

10-04-05, 01:33 PM

I've seen one drive take out another until the failed drive is replaced on the same channel. However, I'm convinced that most of the data lost when this occurs is because it is almost impossible to tell which drive actually failed without taking both drives out and placing them in a different machine to diagnose, pick the wrong one and there goes the array (ask me how I know). The thing I like about the scheme that we are looking at is even if two drives are actually out you only loose the data on those two drives. I've suffered a multi-drive failure on a 1 TB Raid5 system (the drives were on different controller channels that time) and I'm more than willing to take the chance of loosing two drives worth of data vs. loosing the entire array. In terms of a performance hit I remember reading a comparison of raid controllers where all the higher end boards put a drive per channel (4 channels) and the less expensive model used only two . Funny thing about the review was that the low end model soundly trounced the big boys in quite a few of the tests to the point that the reviewer wondered if something was wrong with the other boards. I guess that my point is that there is no perfect storage system that will ensure that no data is ever lost and at this price point I haven't seen a better thought-out system.

Thomas J. Coyle

10-04-05, 03:20 PM

Hi all,

For those of you who want to roll your own server, but use Lime Techs s/w package, there are presently 30 of the Intel D865GLCL motherboards for sale on eBay by www.spartantech.com for $117 plus shipping. Newegg.com is out of stock and mwave.com does not carry the motherboard. None of the local shops in the Los Angeles area carry this motherboard as far as I can tell. I could only find one reseller on the web that was offering this board for sale. I think that I might grab one of the motherboards off of eBay just incase there might be issues getting the s/w to work with other motherboards that use the Marvell 88AE8001 Gigbit controller chip.
Regards,
TCIII

ohlwiler

10-04-05, 03:45 PM

If you are looking for the Intel D865GLCLK motherboard check Zipzoomfly, you will be pleasantly surprised.

I ordered a motherboard, processor and the software today.

Fingers crossed,
Scott

Thomas J. Coyle

10-04-05, 04:10 PM

ohlwiler,

According to the Zipsoomfly listing, the D865GLCLK is on back order and is selling for $129.99. Did I miss something?
Regards,
TCIII

J. L.

10-04-05, 04:24 PM

ohlwiler,

According to the Zipsoomfly listing, the D865GLCLK is on back order and is selling for $129.99. Did I miss something?
Regards,
TCIII
Never used "zipzoomfly: or purchased from them, but I put D865GLCLK in their search field and it came back with two matches, an out-of-stock retail boxed version as you found and an in-stock "bulk" version for less than 100. (Not sure what is not included in the "bulk" version, but there might be some difference other than the glossy printed box)

In either case, since I ordered the entire setup from Limetech, I don't need to buy any additional motherboards from anyone else... they are all yours... :D

ohlwiler

10-04-05, 04:26 PM

The major difference is the warranty, but maybe some of the extras too. Newegg also shows a "bulk" version with some extras included. All I really want is the backplane cover thingy.

Scott

jalaram

10-04-05, 04:27 PM

ohlwiler,

According to the Zipsoomfly listing, the D865GLCLK is on back order and is selling for $129.99. Did I miss something?
Regards,
TCIII

Directron has it for 109 here (http://www.directron.com/d865glclk.html). Shipping came to 7.90 for me to NH.

Thomas J. Coyle

10-04-05, 05:02 PM

To all,

Apparently a whole lot of Intel D865GLCLK bulk packaged motherboards (1 year warranty) got dumped on the surplus market. The lowest eBay price is $87 with free USPS 5-10 day shipping. Zipzoomfly.com has an OEM bulk price of $94 with free 2nd day shipping. Since I live in CA and Zipzoomfly.com is located in Newark, CA, I might just go with them rather than the seller on eBay.
Regards,
TCIII

Birdman71

10-04-05, 05:28 PM

Another software question:

Is it possible to exclude any of the drives from the array and run 1 or 2 drives as stand alone?

For media storage I can imagine a need for high-speed writes (HD recording across network) and it may be beneficial to allow at least 1 drive and maybe 2 to operate without the parity calculation and write. I may be way off base here but it seems that there could be a significant performance advantage, especially if multiple writes are being performed to different volumes simultaneously.

Ultimate flexibility would be on a per drive assignment, either free-standing or as part of the parity cluster.

Birdman

limetech

10-05-05, 02:17 AM

limetech

10-05-05, 02:26 AM

The major difference is the warranty, but maybe some of the extras too. Newegg also shows a "bulk" version with some extras included. All I really want is the backplane cover thingy.

Scott

The "bulk" packaging includes:
- the motherboard itself
- the cpu heatsink bracket (some assembly thus required)
- an unmarked CD with various Windows drivers
- a single 80 pin IDE cable
- a floppy cable
- the backplane cover thingy
- a nice white box

limetech

10-05-05, 02:29 AM

limetech

10-05-05, 02:45 AM

Tom, you know far more about the subject than I do but I will have to disagree. I've had 2 different RAID5 solutions where, when the Master (or IDE0, or whatever) went bad it caused the Slave (or IDE1) to ALSO be marked bad. This isn't misinformation. The Slave wasn't bad obviously, and had to be reinitialized. But it can cause a real headache. This was with both Software RAID5 in Windows 2000 and with a competitor product. And both times the support people I talked to knew immediately what had happened and how to fix it. So... I'm not sure this is an issue with your solution or not. But it definately does happen, and is a concern.

Concern noted. My point was simply that there is no interdependence between the two devices. In the case you cite here, had it been the Slave which "went bad" probably the master would have also been marked bad (by an obviously flawed driver stack).

Yes, this is an issue with multiple device on a single channel or bus. I've seen SCSI systems where a bad device caused the whole bus to hang. But in this case, the driver was smart enough to not fail all the devices.

In the UnRaid driver we NEVER disable more than one device. For example, if a drive fails in such a way as to hang the channel, then what will happen is the low-level IDE driver will fail all the queued I/O to it. But when UnRaid sees the first write error, it will fail that drive (that is, disable it). Thereafter, any other error results in just failing the original host request.

In this case what would happen is your windows application using the share would see an i/o error. The UnRAID management console will show a disable disk and possibly a non-zero error count on other disks. I would guess that 99% of the time the correct disk would be marked as failed. You would then remove this disk and all should be ok again.

This is how it should work by design - but we haven't encountered a disk which failed like this, and we haven't tried to simulate it [add another item to the todo list :rolleyes: ]

limetech

10-05-05, 02:55 AM

limetech,

I checked the Intel website for the indicated motherboard part number. Intel has the following generic part number descriptions on their website:

D865GLC: micro ATX

D865GBF: full size ATX

I assume that adding the additional "L" to the part number adds the optional LAN capability. Intel indicates that the LAN option can be either 10/100 or 1000. Does the "K" term indicate the LAN speed as they do not indicate anything in the board description?
Regards,
TCIII

I haven't been able to figure out Intel's model numbering scheme, but I believe that's correct. You definitely want to get the one with the "LK" at the end of the model number - that indicates it has the PRO/1000 GigEthernet chip.

BTW - I really really really like Intel motherboards for these reasons:
1. Their quality and price is as good as anyone else
2. Their documentation is outstanding. This is reason enough to exclusively use them.
3. Their tech support is actually helpful. Early in this project we noticed a flaw in the bios where, after setting it for USB boot, if you added or removed a hard drive the bios would "revert" to hard drive boot. I sent them an email detailing this "bug" and got a reply that they would "look into it". Well, in a later bios release guess what? They actually fixed it! Don't know if was because of my email, perhaps they knew about this already, but that is true support.

limetech

10-05-05, 03:15 AM

Madpoet,

It's that master/slave failure that people have talked about in the past that has prevented me from rolling out an array in the past. After reading Tom's post I wonder if this issue isn't a problem in regards to their software. Maybe (pure speculation) RAID 5 is the key ingredient to the master/slave problem? If Raid 5 has to write data to every drive in the array for every write event, maybe a failed drive (master or slave) on a channel prevents the other same channel drive from writing (due to write caching). If the array controller senses that both drives had write failures it may flag them both as failed? If this is the scenario than the potential still exists in the unraid softare, but only on the shared parity drive channel. Since the main data drives are treated independently the system would never try to write to a maste/slave config for the other ten data drives.

Again, pure speculation on my part. Hopefully someone with a better understanding of this will chime in.....inquiring minds want to know.
Very good analysis.

I'm pretty excited about this application. I've been following the NAS threads under the media server forum and like the expandability options here with 12 drive slots v. 4.
Any discussion of UnRAID over there?

Tom,
What's the upgrade path from PATA to SATA for people that purchase now?
Do you mean for those who purchase a complete server? If so, you could take it apart and replace some/all of the disk trays with SATA versions along with replacing the PCI disk controller cards.

If you mean for the software-only product, then SATA support will be a free upgrade.

Also looking for more information regarding the mixing of PATA and SATA in a system once you have the SATA version finished. How many SATA ports are on the MBs you are recommending and are you planning on supporting using both PATA and SATA MB channels at the same time?

Birdman (still trying to drum up support for a powerbuy)

There's no fundamental reason why you couldn't build a system with a mix of PATA and SATA drives. You would have to decide how many of each type you want and their slots. It's a bit complicated creating the correct "configuration" file that tells the UnRAID software which devices correspond to which slots - but definitely do-able.

I don't anticipate offering any preconfigured products which mix PATA and SATA, but you could certainly roll your own.

justmikie

10-05-05, 10:58 PM

3-way

10-06-05, 04:40 PM

musicmann

10-06-05, 05:58 PM

I have 4 PATA and 8 SATA drives right now. I didn't realize I could put them both in the same box / system. Would loading all 12 drives in a single system be hot? Would one PATA controller card and one SATA controller card cover me?

I'm thinking you'd go 0 PATA controller card and 2 SATA controller card. For the PATA, you'd use the on-board connections, and each of the SATA cards would give you 4 connections. (Assuming no on-board SATA).

Then again, I'm guessing ...and trying to get my number of posts up.

Thomas J. Coyle

10-06-05, 08:31 PM

ohlwiler,
Have you received your motherboard and processor from zipzoomfly.com yet?
Regards,
TCIII

ohlwiler

10-06-05, 10:29 PM

I received my motherboard today. The motherboard came with the back plane cutout, floppy cable, ide cable and driver CD. Testing will have to wait until my case arrives next week. It was my first time ordering from Zipzoomfly. I usually buy from Newegg. Everything went smooth.

It looks like the software from Lime may be the last item to arrive.

Scott

leonowski

10-07-05, 12:04 AM

Tom,

Would something like this work in the IDE version of Un-RAID?

http://www.firewire-1394.com/ide-sata-hard-drive-adapter-bridge.htm

This could probably allow people to run SATA and IDE drives mixed together.

-Noel

*Update*

OOPS, it says Windows OSes only. I guess it needs drivers. Nevermind.

limetech

10-07-05, 01:33 AM

I must be a lot less curious than the average AVSForumer. You guys are keeping Tom way busy with questions :) . Being the impatient type I just went ahead and bought one. I installed 3 500gb drives when I first got it. Installation went fine and I had a terabyte of space to fill. I started ripping movies and realized that I needed more space (I have 600 movies), had some "extra" paypal money and ordered 2 more 500gb disks. When I went to install them, I couldn't see the utility to stop the array (I could , however, see the files on the 2 drives).
One thing we need to work on in the s/w is better feedback about what's happening during an operation. Some operations take "a long time" and it can appear that the system is locked up when in fact it's still crunching away. For example, to "mount" a reiserfs file system it takes close to 45 seconds once the disk capacity is over around 200GB. Now when there's multiple drives to mount we try to do them in parallel, but it still takes almost 45 sec times the number of drives to complete (seems to be highly CPU-bound). Anyway, my point is that sometimes, especially when adding drives, it's possible to get into a state where it appears things have ground to a halt. I'm not saying this happend in your case, but other people have, shall I say, pointed this out to us as a problem to be addressed.

I looked on the website for help for my problem and didn't find any. It was Saturday afternoon, and my 11 year old daughter was expecting to watch a movie that night, so I panicked and shut off the power supply. When I powered back up, I couldn't see the utility or the files on the drives. Now I was in trouble. I sent Tom and email, and an hour later he called me and quickly fixed my problem. Great service, especially on a Saturday.
I also have a 9 year old daughter who expects to watch her movies when she's been told she can, and is very upset when that can't happen...........

The only downside was that it took nearly 3 hours before the array was back up. I remember the website saying something about 400gb in 45 minutes....

The system was clearing your new drives before adding them to the array. The current code executes this "clearing" operation before starting the array because it was simplest and safest to do it this way - but we'll fix this in another release. Another way to add new drives is to first "initialize" the array configuration data (via Tools menu), in effect making it think all the drives are "new". Then when started, the shares become available much faster and a backgroun operation is started to regenerate parity. A problem doing it this way is there's no recovery if a drive fails during this operaiton.

For my first performance test, I tried playing a movie and ripping one at the same time, and the playback stutters. However, playing 2 movies from the same disk at the same time worked fine.
I'd be interested in knowing some more details about your first test:
a) was ripping done on same PC as viewing?
b) was ripping writing to same disk as the one being read for viewing?
c) what's your network configuration and speed?
d) anything else happening on your network?

p.s. Compusa had 500gb on sale for $309, with a $40 rebate so I bought 2 more... it's a disease :o
That's a great price, but man, I hate dealing with rebates....

limetech

10-07-05, 01:39 AM

I have 4 PATA and 8 SATA drives right now. I didn't realize I could put them both in the same box / system.
We haven't actually tried this, but it should work ;) Note that our preconfigured server is only all-PATA or all-SATA. You would have to build your own system and use our s/w.

Would loading all 12 drives in a single system be hot?
You definitely need proper cooling; however, the system will generally run cooler than you might expect because most of the time most of the drives can be spun down.

Would one PATA controller card and one SATA controller card cover me?
You can use the 2 on-board PATA controllers for 4 PATA drives and either a single 8-port SATA controller or two 4-port SATA controllers. We're currently testing a couple 4-port controllers.

limetech

10-07-05, 01:40 AM

...
Then again, I'm guessing ...and trying to get my number of posts up.
Me too :D

limetech

10-07-05, 01:44 AM

...
It looks like the software from Lime may be the last item to arrive.

cott

All the Flash orders are being shipped tomorrow (Friday). You will get a tracking number and email. I apologize for the delay. When we only offered complete servers it was sufficient and desirable to have Friday be "shipping day" - and that is when the FedEx pick up occurs for us. This week, all the other orders will also get shipped tomorrow but next week we'll start shipping within a day of receiving the order.

Thomas J. Coyle

10-07-05, 09:34 AM

Tom,

What is your s/w licensing policy going to be? Will we have to buy a license for each machine that the s/w is on or will you allow more than one up to a reasonable amount? Maybe a discount for additional licenses?
Regards,
TCIII

Thomas J. Coyle

10-07-05, 09:39 AM

ohlwiler,

Thanks for the update. I noticed that the D865GLCKL was back in stock at newegg.com, but the price is substantially more than zipzoomfly.com so I ordered from them. Should have the parts on hand by Tuesday of next week. I plan to run the s/w on both my Asus A7N8X-E Deluxe and the Intel motherboard as I have two DVD servers at the moment. I have asked Lime Technology what their licensing policy is.
Regards,
TCIII

kchad

10-07-05, 12:46 PM

Tell me if I have these scnearios correct.

Time goes by and all drives have spun down.
I read a file from one drive – only the drive I want to read from spins up and the file is read.

Time goes by and all drives have spun down.
I write a file to one drive – ALL drives spin up to calculate parity, the file is written to the target drive, and the parity info is written to the parity drive.

razi

10-07-05, 01:08 PM

Hi,
Given that most of my storage is on external USB 2.0 and Firewire devices, is it possible that your software would work with these external drives. All the data is 'static' i.e it never changes once it has been put on an external drive. Can I run the S/W in offline mode. By this I mean, can I just calculate parity once and be done with it until I need to add an additional drive to the array? This way the parity drive can be off unless needed to either restore data or recalculate parity.

J. L.

10-07-05, 01:15 PM

I think I can answer these
Tell me if I have these scnearios correct.

Time goes by and all drives have spun down.
I read a file from one drive – only the drive I want to read from spins up and the file is read.
Correct.

Time goes by and all drives have spun down.
I write a file to one drive – ALL drives spin up to calculate parity, the file is written to the target drive, and the parity info is written to the parity drive.Not correct.

No need to spin up all the drives since it is easy to recalculate the required parity by reading the existing contents of the blocks on the parity drive and the drive being written to and comparing their old values to the new data being written. We do need to read the old contents of the parity drive and the data drive for those blocks being written to get their old values first for the comparison.

Therefore, only the parity drive and the one data drive need to spin up.

At a single bit level, if the bit on the parity drive was a "1" and the old data bit was a "0" and the new data bit value also a "0" then the value of the parity bit does not need to change since if the number of "1s" at this bit postion across all the drives did not change. It would also not need to change if the old and new data bit values were both a "1".

If however, the old data bit was different than the new data bit (0 to 1, or 1 to 0) then the value of the parity bit would also have to change (from whatever it was to the opposite value) to keep an even number of "1s" across that bit position in the unRaid array. In this example it was originally a "1" so it would have to change to a "0".

Joe L.