or Connect
AVS › AVS Forum › Video Components › Home Theater Computers › Mfusick's How to build an affordable 30TB Flexraid media server: Information Requested.!
New Posts  All Forums:Forum Nav:

Mfusick's How to build an affordable 30TB Flexraid media server: Information Requested.! - Page 23

post #661 of 3344
Don't know why you are being so defensive MFusick.

I posted some data from another forum showing the time it takes to transfer what is the equivalent of a few bluray rips. I doubt that a drive with higher density platters is going to significantly change the outcome (but I could be wrong).

There is an easy way to test the theory. Just get out your stopwatch. That's what AV Science is all about. I have a bunch of different drives and all are fantastic for storage and playback.
post #662 of 3344
Thread Starter 
Some thoughts on hard drives (opinions)

RED WD HDD's certainly has a nice niche in it's targeted consumer category. If I was building a low power pre-fabricated NAS box WD RED drives would be near the top of my list because of cool and quiet operations, long warranty, and great vibration resistance. I think they are idea for that application, and not ill suited for home media servers either. The longer warranty and street prices near WD greens, newer design (1TB platters) and higher performance make them a good choice over WD GREEN- especially since WD recommends them for those purposes and does not recommend GREEN. I never suggested or pretended otherwise. I'd consider RED WD drives personally if I found them under $40 per TB, or on par with the pricing of the better Seagate 7200.14's.

I think your missing the point I am making that the Seagate is a do it all drive with clear advantages in performance that's not well represented by Asassin or in the intention of this thread; Even more is that Seagate also holds a significant advantage in price which is usually paramount in any purchase decision and they are not at all ill suited for use in a HTPC, PC, mediaserver or NAS box- Making them highly attractive.

I understand completely WD's product strategy. They have differentiated and made specific certain product lines for certain types of applications. This "specialized" approach makes sense if your looking for certain attributes in your drive for very distinct applications. Example being your right about the RED being a fine product for it's intended purpose. But- I am not sure I see the benefit or advantage to this approach. I think it hurts WD and helps Seagate. Seagate uses a one drive for all approach that is very different. It's a good balance on most characteristics, and it's got economies of scale in it's favor.

Here is how I sum up my understanding of the WD line:

Black - High performance. Energy pigs. Equivalent of an older muscle car with a modified supercharged V8 engine. Not going to deliver good gas mileage, and not as efficient in design as newer models- but excellent at it's specialized area. Like a muscle car is good at acceleration, power and speed- A black drive is too- but in seek times. They have overbuilt cache, boards, use 7200rpm and place seek time as a priority. This used to make them great OS drives before SSD stole that market. Today they are still good at seek times, but really do not hold any advantage in sequential speeds like copying a big movie file, making them terrible choices for media servers, and HTPC's. A Seagate 7200.14 is faster in comparision. The high price, high heat, high noise, and high energy consumption makes this drive a poor choice for most in this forum.

Blue - Used to be the bread and butter of WD. The normal 7200rpm drive. Lost much of its thunder these days. Market shifted in 2010 towards the lower heat, noise, and better energy profile of the green 5400rpm drives. In 2010 when we were at 500GB platters- the GREEN drives debuted as newer more efficient designs that could compete on performance, and win on energy profile. Green also had aggressive pricing. Most switched from the older blue's to the newer greens - and rightfully so. Blue has not regained it's former glory in the current day market place.

Green Debuted as an energy saving, low cost, low heat and low noise drive. Found good success in basic storage applications where it's benefits were appreciated. (HTPCS's biggrin.gif) Remains a great choice for basic storage in a PC or HTPC today for the same reasons it always has been. But if we peel away the layers we'll see that 2 years ago Green drives had a higher platter density, newer more modern design and could hang with the 7200rpm drives of yesterday in performance while being better at energy and selling cheaper. Today, opposite is true. They are less advantaged because new 7200rpm drives with even more modern designs and higher density have hit the market taking back the performance advantage and also the price advantage. Two big blows to the viability of Green drives; also the newer 1TB platters and more efficient designs have allowed 7200rpm drive to operate with a better energy profile too making them even more attractive. Green drives are basic storage specific and are not recommended for media servers, NAS boxes or many of the duties storage drives are used for today. The market has shifted from internal bare HDD storage to NAS- with NAS boxes and home servers which typically hold many HDD's. Green drives do not have advanced anti vibration technologies like RED making them a poor choice in those applications. They also have a shorter warranty which sucks. They suffer from TLER issue and should never be used in RAID set ups. Lastly, it is a controversial topic but the green drives have a few power savings features that lower total reliability- specifically in 24/7 applications like NAS boxes and Servers. For this reason- WD does not recomend green drives for those purposes and the change in the market to those purposes was the main reason for the introduction and design of the RED drive line that is recommended for those purposes. The controversial issue the agressive settings in the "intellipark" head parking feature that parks the heads on a green drive after 8 seconds. The idea is to shut down the drive quickly when not in use and save on energy consumption; But this feature has a side effect in 24/7 applications like NAS boxes, File servers, and linux based machines. The side effect is the drive gets subjected to an excessive amount of ON/OFF cycles which causes premature wear and lower endurance. It's not going to die instantly if you use a green drive in a NAS box, linux box, Unraid/ubuntu or a media server- but it's probably not an ideal use for them per WD and MFG recommendations. You can disable this feature with WDidle8 or WDidle3 firmware hack which also removes the "click" of the head parking which is nice for internal HTPC applications if it bothers you. The click is not loud, but it can be annoying.

Here is some info on the WD GREEN head parking and reliability issues:

http://community.wdc.com/t5/Desktop-Portable-Drives/Green-Caviar-High-Load-Cycle-Cout-after-short-operation-time/td-p/15731

http://community.wdc.com/t5/Desktop-Portable-Drives/Green-Caviar-High-Load-Cycle-Cout-after-short-operation-time/td-p/15731/page/16

http://techreport.com/forums/viewtopic.php?f=5&t=78891

http://www.networkedmediatank.com/printthread.php?tid=20686

http://wdc.custhelp.com/app/answers/detail/a_id/5357/session/L3RpbWUvMTM2MjE1MjE0Ny9zaWQvQktRUmI0a2w%3D
Straight from WD themselves:


If you wanted to know more there is tons more information here: http://lmgtfy.com/?q=WD+Green+head+parking+issue


*disclaimer- My goal is not to disparage or denounce Green drives as I believe they are good choices for their intended and MFG recommended purposes; I just want create some awareness on the specificity of their intended purpose and possible complications of using them outside that purpose.



RED The newest product line WD offers with a specialized design and recommended purpose for NAS boxes, Mediaservers, 24/7 operations. They do not have the TLER issues of green drives making them much better for these purposes and for any RAID set ups. They are designed and certified for use in server cases, and cages with >5 HDD's and feature advanced anti vibration technologies making them more appropriate solution in such configurations. RED uses the same 5400rpm rotational spindle speed of the GREEN drives making them very attractive on energy profile. RED uses 1TB platters and is a newer design so it's normal to see both energy figures and performance figures superior to Green drives. RED is basically the evolution of the green drive into today's market that has a strong need for NAS, 24/7 media servers and applications that are more common than a few years ago- RED offers a longer warranty which is nice too. They still have the 5400rpm speed so the performance is acceptable but below what you see from a modern 7200rpm design with 1TB platters, making them a good low performance solution.

RE and AVG/P Expensive to the point I'll not make much of these two lines. WD makes a audio/video drive with wear leveling and high endurance for use in DVR and constant video recording applications and also makes many enterprise level drives that do not have head parking or TLER issues and are designed for high endurance in demanding enterprise applications like servers and data RAID set ups. The price of these drives makes them a poor choice for HTPC's and cheap home storage.

That is basically how I'd sum up my personal understanding of WD lines- but I for sure appreciate the singular and simplistic approach of Seagate by comparision.

Seagate has designed a one size fits all drive that tries to balance everything into a very attractive product. Enter the 7200.14. It's a modern design with (3) 1TB platters making it both higher performance AND more energy efficient. It does not have any head parking issues or TLER issues. It's a good solution for internal PC and HTPC storage. It's a great solution for home media server or NAS too. It's faster than 5400rpm drives, it's on par and competitive with energy (something 5 platter older 7200rpm are not) and because of the one size fits all approach it benefits from significant economies of scale and aggressive market pricing making it the highest value HDD. It's typically the lowest cost per TB today ($40 per TB).

Toshiba has a 3TB that has some buzz and the Hitachi drives are very good but use more energy.
Edited by Mfusick - 3/4/13 at 12:57pm
post #663 of 3344
It's completely obvious you don't know who I'm talking about. Google is not going to help you with this one.
post #664 of 3344
Lets talk about other things now besides speed, shall we?

Since there are 1TB platters already, and HDD manufacturers are able to fit five platters in a drive, why haven't we seen 5TB drives yet? I doubt cost is an issue.
post #665 of 3344
Thread Starter 
Quote:
Originally Posted by amarshonarbangla View Post

Lets talk about other things now besides speed, shall we?

Since there are 1TB platters already, and HDD manufacturers are able to fit five platters in a drive, why haven't we seen 5TB drives yet? I doubt cost is an issue.

Yes.

Your exactly right.

Your going to see 5 platter (1TB platters) and 5TB drives soon. It's next logical step. 4TB is 5platter with 800GB platters, which is why I prefer the 3TB's.

Usually a new design like the 1TB platter comes out and hits a product line (like the 3TB Seagates) and once production ramps up, reliability is proven and economies of scale ramp up it finds it's way into more products, and replaces older technologies.

Example:

1TB drives used to have dual 500GB platters, but modern designs use a single platter and 1TB density.

It's certain to happen. Just not yet.

I think a 5TB drive with (5) 1TB platters would be beasty for storage. 10TB from two speedy drives. I like it.
post #666 of 3344
Quote:
Originally Posted by Mfusick View Post

Yes.

Your exactly right.

Your going to see 5 platter (1TB platters) and 5TB drives soon. It's next logical step. 4TB is 5platter with 800GB platters, which is why I prefer the 3TB's.

Usually a new design like the 1TB platter comes out and hits a product line (like the 3TB Seagates) and once production ramps up, reliability is proven and economies of scale ramp up it finds it's way into more products, and replaces older technologies.

Example:

1TB drives used to have dual 500GB platters, but modern designs use a single platter and 1TB density.

It's certain to happen. Just not yet.

I think a 5TB drive with (5) 1TB platters would be beasty for storage. 10TB from two speedy drives. I like it.

Yup. Yummy indeed. You will probably get about 9.2TB usable space after formatting.

Besides that, hard drive prices haven't still stabilized from the Thai flood. Prices aren't coming down like they used to. 4TB drives should have been $120 by now. Not sure if this will delay the 5TB drives. It's all ultimately bad for us customers.
post #667 of 3344
Thread Starter 
I think you'll see $99 3TB and $199 5TB. 4TB probably in between ($150)

In any case the 3TB is still the better buy and choice IMO. Cheaper per GB, and probably higher performance/lower energy.

Even if the 5TB outperforms the 3TB, it's going to cost $50 a TB most likely because it's new. 3TB will benefit from economies of scale and it should be a better value at $33/TB.

The new 1TB platter drives like the 7200.14's are going to be around for a while. The 1TB and 2TB seagates with the modern design and 1TB platters are also nice drives, but I imagine the size of the 3TB and low cost make it the best choice for most storage applications.
post #668 of 3344
Quote:
Originally Posted by Mfusick View Post

I think you'll see $99 3TB and $199 5TB. 4TB probably in between ($150)

In any case the 3TB is still the better buy and choice IMO. Cheaper per GB, and probably higher performance/lower energy.

Even if the 5TB outperforms the 3TB, it's going to cost $50 a TB most likely because it's new. 3TB will benefit from economies of scale and it should be a better value at $33/TB.

The new 1TB platter drives like the 7200.14's are going to be around for a while. The 1TB and 2TB seagates with the modern design and 1TB platters are also nice drives, but I imagine the size of the 3TB and low cost make it the best choice for most storage applications.

Aahh $99 3TB drives will be a dream come true. I doubt that's happening before the year's end.
post #669 of 3344
Thread Starter 
WHy not ?

I bought 5 of the Seagates for $89 each on black Friday last year.

U mad ??biggrin.gif
post #670 of 3344
Quote:
Originally Posted by Mfusick View Post

WHy not ?

I bought 5 of the Seagates for $89 each on black Friday last year.

U mad ??biggrin.gif

post #671 of 3344
Thread Starter 
post #672 of 3344
Why am I getting over 2GBps reads on my FlexRAID pool o_o. I understand FlexRAID can read the drives in parallel but 2GBps+ speeds are ridiculous!!! Could you post benches of your FlexRAID array please? Assasin, would love to see some benches of your pool also.


Quote:
Originally Posted by Mfusick View Post

Warning: Spoiler! (Click to show)

Lol
post #673 of 3344
Thread Starter 
Your bench is not making any sense to me. Can you explain it ?

It's faster than my SSD C drive:



And I'm stuck at the Sata 3 interface limits eek.gif
post #674 of 3344
Quote:
Originally Posted by Mfusick View Post

Your bench is not making any sense to me. Can you explain it ?

It's faster than my SSD C drive:
Warning: Spoiler! (Click to show)

There's nothing to explain. Just chose my FlexRAID pool and ran the bench. Got those crazy number! You should try benching your pool too, see if you get similar numbers.
post #675 of 3344
Thread Starter 
Did you bench a RAM disk or something ?

Your bench appears quite a bit faster than mine. To the extent it makes no sense.

I'm tapped out at 1100MB/sec because Sata3 only does about 525/MB sec.

Your bench is just outrageous. You must have it cached with a RAM disk would be my guess. I need explanations.
post #676 of 3344
Quote:
Originally Posted by Mfusick View Post

Did you bench a RAM disk or something ?

Nooooo, it's my FLEXRAID POOL. You can see the letter V: in there.
post #677 of 3344
Thread Starter 
Quote:
Originally Posted by amarshonarbangla View Post

Quote:
Originally Posted by Mfusick View Post

Did you bench a RAM disk or something ?

Nooooo, it's my FLEXRAID POOL. You can see the letter V: in there.

Something is wrong.
post #678 of 3344
Quote:
Originally Posted by Mfusick View Post

Something is wrong.

Is your FlexRAID pool up and running? Try and bench your array...
post #679 of 3344
Also
Quote:
Originally Posted by Mfusick View Post

Bought some server parts today for another build for a friend:

Norco 4220
Asrock Z77 Extreme 4
G.Skill 8GB DDR3
i5 3570 CPU
Seasonic 620watt PSU
Seagate 7200.14 3TB's
Samsung 840 128GB SSD
Norco SFF8087 to 8087 cables
Norco SFF8087 reverse breakout 4 Sata
3 pin Y fan cable splitter for using mobo headers for internal fans
Norco 120mm optional fan blade bracket
Xigmatek eXTREME Silent FCB 120mm and 80mm fans replace stock Norco OEM fans
LITE-ON slim internal DVD burner drive for slimline Norco bay
Intel 10/100/1000 PCI-Express Lan card 1XRJ45
WHS for OS

Going to be another nice build.

I just need to find a deal on the IBM M1015 card again. Any advice or help ?
Ebay is dried up at the moment.

I'm probably going to do server build thread for this one and do a better job archiving my steps from beginning to end.

You have bought the parts already but, have you looked at the Xeon E3-1230 V2 CPU? Has hyper-threading and fairly power efficient too.
http://www.newegg.com/Product/Product.aspx?Item=N82E16819117286

Thinking of getting this myself for my server.
post #680 of 3344
Thread Starter 
A solid choice for sure.

My flexraid array does not populate in my ATTO from my desktop. Then again it would be over LAN so it would be limited by that. I'd have to run it locally to make it worth anything.

I'm too lazy to click remote desktop. It's late. lol. Remind me another day and I'll run it for you.
post #681 of 3344
Quote:
Originally Posted by Mfusick View Post

A solid choice for sure.

My flexraid array does not populate in my ATTO from my desktop. Then again it would be over LAN so it would be limited by that. I'd have to run it locally to make it worth anything.

I'm too lazy to click remote desktop. It's late. lol. Remind me another day and I'll run it for you.

Lazy bum sadfaceu.png
Warning: Spoiler! (Click to show)
Kidding.
post #682 of 3344
Thread Starter 
Well I knew your results were bogus. lol. You know how fast that is ?

I was certain I could beat you so I actually posted a reply the first time you posted it with

" I RAISE YOU: "

And posted my benchs, then realized mine were slower and deleted my post in shame.. lol.

But it was excessive enough to the point I knew it was bogus. The only way I see hitting those speeds is a ram disk cache.

My C drive barely hits 1100MB/sec, and it's probably much faster than most in this forum.

So- explain where did the bench come from ?? I am undeniably interested.
post #683 of 3344
Quote:
Originally Posted by Mfusick View Post

Well I knew your results were bogus. lol. You know how fast that is ?

I was certain I could beat you so I actually posted a reply the first time you posted it with

" I RAISE YOU: "

And posted my benchs, then realized mine were slower and deleted my post in shame.. lol.

But it was excessive enough to the point I knew it was bogus. The only way I see hitting those speeds is a ram disk cache.

My C drive barely hits 1100MB/sec, and it's probably much faster than most in this forum.

So- explain where did the bench come from ?? I am undeniably interested.

Do you have teamviewer? I will run the bench right infront of you...
post #684 of 3344
I think you are just seeing cached read/write test results. I'm no expert on Atto benchmark but the file sizes you are using are so small that they will produce cached results. Caching can appear in any level of your system (drive, controller card, operating system). For instance a Linux os will use all available memory for caching, Windows does similar things.

I spent quite some time back in 2009 trying to find a suitable benchmarking tool which would provide results comparable to real-world scenarios. My main objective was to find out if configuration options (like NCQ) made any difference to actual performance. I ended up using IOZone since it was the only tool easily available on both Windows and Linux platforms. A secondary objective was to find out how the performance changes from local to over-the-network access. Since I had a mixed client environment, I wanted to know what is theoretical maximum performance (=local) and compare it to performance on different Windows machines to see if I had a problem somewhere.

One thing I learned from the whole experience is that nothing beats real live testing. Synthetic tests are nice for but if you want to really know the difference on your environment you will have to do testing yourself. I made a simple batch script which copied a given source directory structure to given target and showed the final transfer time/speed. And there is nothing wrong just using the transfer speed reported by OS. That's basically all that counts in htpc storage. If you were to talk about storage for databases or virtualisation benchmarking would be far more complex.
post #685 of 3344
Thread Starter 
Your probably right. I'm curious to try mine at home now. (At work now)
post #686 of 3344
Thread Starter 
Help me with something: (reposted from Assassin thread)

1TB = 100GB = 1048576MB

1048576 MB divided by an average Seagate speed of 155 MB/sec speed is 6765 second (112 minutes)

1048576 MB divided by an average WD GREEN speed of 95 MB/sec is 11037 second (184 minutes)

So it's well over an hour shorter time in the actual time difference for a 1TB task to complete. I basically just used the Tomshardware average benchmark for each drive since it's most like what I see myself in real life.

That means your server is done with a 3TB parity calculation 3.5 hours sooner with a Seagate eek.gif I'm imagining the drives spinning down and/or remaining idle for an additional 3.5 more hours than the GREEN drives would actually make the Seagates more energy efficient in the real world application of a media server running Flexraid. Not sure, just a guess. Would be an interesting subject to explore if the higher performance of the 7200rpm drives actually leads to an additional costs savings when your factor in the time value saved or not.

I am not exactly sure how it all plays out but I'd imagine a drive running at half the power because it is @ idle for 3.5 hours more would end up being more energy efficient and conserving more energy than a slower drive running for an additional 3.5 hours to calculate 3TB of parity.


My understanding is that Flexraid reads all drives simultenous and any serious CPU can keep up with the math and parity calculations easily. So your time it takes is mostly dependent on the time it takes to read all your data, or write the parity data. The main determining factor being the slowest or weakest link. So if you 5 small or empty HDD's and 1 full 3TB it's still going to take the same time as if you had 5 full 3TB assuming all 3TB read or write at the same exact speed. Your largest, Slowest most full drive - or whatever real world combination leads to the slowest time is basically the total time it will take.

So if you have a mix of 5400rpm Green and 7200RPM Seagate your probably limited by the green drive and there would be no cost savings or quicker completion. (me now). But if you had 100% all 7200rpm faster drives then it's taking less time. That means less time for each update, Less time for each validate, and less time for each verify process. It's probably not a factor unless your comparing a totally new build using 100% faster or slower drives versus each other and I'm guess many are like me with a mix of HDD's. But- interesting thought that the higher performance cheaper HDD is actually saving your even more electricity because of the better performance it has.

I think a full tilt 100% going drive uses 10 watts, idle about 5 watts and spun down perhaps 1 watt. I'm not certain though.
post #687 of 3344
Quote:
Originally Posted by Mfusick View Post

My understanding is that Flexraid reads all drives simultenous and any serious CPU can keep up with the math and parity calculations easily. So your time it takes is mostly dependent on the time it takes to read all your data, or write the parity data. The main determining factor being the slowest or weakest link. So if you 5 small or empty HDD's and 1 full 3TB it's still going to take the same time as if you had 5 full 3TB assuming all 3TB read or write at the same exact speed. Your largest, Slowest most full drive - or whatever real world combination leads to the slowest time is basically the total time it will take.

I don't think it's really worth making lots of assumptions about

You are correct that the bottleneck has to be either calculation, read i/o, or write i/o. I don't think spectrum is going to divulge the way that flexraid reads block data from your HDDs in snapshot mode (and I don't think he should be expected to), but I would guess there is some optimization going on since people claim much shorter parity build times when all drives are empty.

Even though I don't think it's worth making assumptions, I will make the assumption that the limiting speed is write i/o on the parity disk
post #688 of 3344
Quote:
Originally Posted by Mfusick View Post

1TB = 100GB = 1048576MB

The middle part of this was incorrect

1TB = 1024GB = 1048576MB
post #689 of 3344
Since you are doing parity updating the speed will be limited by the write speed of parity drive. I'm also quite certain that you will not see real-life figures near the 150MB range for the simultaneous read on multiple drives with processing (parity calculation) taking place. But this you can easily test on your own system. On my UnRaid system I'm getting ~75MBps for parity checks for a mix of Samsung drives with ~110Mbps lowest max transfer speed. This with an even spread of disks on different controllers either on motherboard or PCI-express bus and only for a check not full-blown update. My bet is that the actual difference between the two drive types in real life usage is very much smaller than difference of their max transfer rates.

All the energy consumption details are easily available for any drive, just dig them out and do the math. My storage system (12 disks) is pulling 65W when ripping a blu-ray. Only two drives (target and parity) are spun up. Most of the time the system is completely idle and drawing ~45-50W. It's not only the consumption itself but also the strain on the PSU when spinning up a lot of disks which I try to take into account. With lower wattage I can live with a bit smaller (=cheaper) PSU. I know you could get away with even much smaller low-stress figures but this is sufficient for me.

Edit: Dark_Slayer beat me to it on the write speed of the parity drive being the limiting factor...
Edited by henris - 3/1/13 at 2:29pm
post #690 of 3344
Quote:
Originally Posted by likelinus View Post

Why do I get the overwhelming feeling this guy is the Alkemyst of this forum? It's rare you see someone with just a superiority complex and ego. Seriously, I read his replies and post and it's just sad.
That's harsh. Nobody beats Alkemyst. biggrin.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Home Theater Computers
AVS › AVS Forum › Video Components › Home Theater Computers › Mfusick's How to build an affordable 30TB Flexraid media server: Information Requested.!