What kind of speeds do you guys get on your FlexRAID pool? - AVS Forum
Forum Jump: 
 
Thread Tools
post #1 of 28 Old 03-03-2013, 09:23 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Getting some ridiculous numbers on my FlexRAID pool. Wanted to see if others are seeing similar speeds. Some here said this was running in the cache. Well, that cannot be possible.

image096f.jpg

image097hk.jpg

I posted this on the official FlexRAID forum expecting an explanation but Brahim claims this is a "secret sauce" which he can't reveal. When I forcefully demanded an explanation and accused him of implementing bogus features, my post was deleted. Something shady is going on here...
amarshonarbangla is offline  
Sponsored Links
Advertisement
 
post #2 of 28 Old 03-03-2013, 11:33 PM
AVS Special Member
 
Nevcairiel's Avatar
 
Join Date: Mar 2010
Location: Germany
Posts: 1,004
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 6 Post(s)
Liked: 107
A FlexRAID pool does not make your drivers significantly faster. It can perform a bit of pre-fetching, but in the overall transfer speed that doesn't matter. There is no striping in FlexRAID, so it can't read one file from multiple discs for speed up (like a RAID 5/6 could)

My guess is that the IO patterns in such benchmarks are either hitting some kind of cache, or FlexRAID detects the benchmark attempt and artificially speeds it up.

If you want a real world test and have a SSD in that system, then run some read/write operations by copying files around from/to the ssd (to ensure the source/destination is fast enough), or if you have a mingw/cygwin environment with the "dd" tool, thats usually also good for simple read/write speed tests.
Nevcairiel is offline  
post #3 of 28 Old 03-04-2013, 04:11 AM
Advanced Member
 
macks's Avatar
 
Join Date: Jan 2013
Posts: 714
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 31
It's writing and reading from multiple discs. Basically his "secret sauce" is providing false benchmark results.
macks is offline  
post #4 of 28 Old 03-04-2013, 05:18 AM
AVS Addicted Member
 
Mfusick's Avatar
 
Join Date: Aug 2002
Location: Western MA
Posts: 21,441
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 62 Post(s)
Liked: 654
Right.

A true test is just copy or paste to or from it - using SSD as the other destination source.

I like terra copy better than the standards windows copy paste function but either one would do.

-

"Too much is almost enough. Anything in life worth doing is worth overdoing. Moderation is for cowards."
Mfusick is online now  
post #5 of 28 Old 03-04-2013, 07:00 AM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by macks View Post

It's writing and reading from multiple discs. Basically his "secret sauce" is providing false benchmark results.

These are regular hard drives. Even if the benchmark was reading from multiple discs at the same time, there's no way they could produce upwards of 2GBps read speeds.
amarshonarbangla is offline  
post #6 of 28 Old 03-04-2013, 07:26 AM
AVS Special Member
 
Dark_Slayer's Avatar
 
Join Date: May 2012
Posts: 2,291
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 51 Post(s)
Liked: 226
Quote:
Originally Posted by amarshonarbangla View Post

I posted this on the official FlexRAID forum expecting an explanation but Brahim claims this is a "secret sauce" which he can't reveal. When I forcefully demanded an explanation and accused him of implementing bogus features, my post was deleted. Something shady is going on here...

Why should he or anyone else be demanded to explain why the benchmarking tool that you chose to use doesn't provide accurate numbers?? How many other benchmarks did you try? How many do you know of? Have you even bothered to just do a large file copy like nearly every one of us suggests?

I don't even know why this was posted in the support forum, but his reply was rather silly. I can't tell if he really has that big of a superiority complex, or (more likely) he was just being cheeky.

Is there a reason you have so much faith in the atto benchmark tool?
Dark_Slayer is offline  
post #7 of 28 Old 03-04-2013, 07:31 AM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by Dark_Slayer View Post

Why should he or anyone else be demanded to explain why the benchmarking tool that you chose to use doesn't provide accurate numbers?? How many other benchmarks did you try? How many do you know of? Have you even bothered to just do a large file copy like nearly every one of us suggests?

I don't even know why this was posted in the support forum, but his reply was rather silly. I can't tell if he really has that big of a superiority complex, or (more likely) he was just being cheeky.

Is there a reason you have so much faith in the atto benchmark tool?

I have seen similar sequential read/write numbers in CrystalDiskMark and Blackmagic Disk Speedtest. I get 80MBps-100MBps in real world tests, which is what I should be getting in these benches, or something close. The astronomic differences between synthetic and real world tests is what's confusing me. There's definitely something going on in FlexRAID that's messing with these benchmarking tools.
amarshonarbangla is offline  
post #8 of 28 Old 03-04-2013, 07:38 AM
Advanced Member
 
macks's Avatar
 
Join Date: Jan 2013
Posts: 714
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 31
Quote:
Originally Posted by amarshonarbangla View Post

These are regular hard drives. Even if the benchmark was reading from multiple discs at the same time, there's no way they could produce upwards of 2GBps read speeds.

I can't see the benchmark you posted(at work). 2GBps or 2gbps would make me think it is in memory. Either way it is a false benchmark that doesn't reflect reality in the slightest.
macks is offline  
post #9 of 28 Old 03-04-2013, 07:55 AM
AVS Special Member
 
Dark_Slayer's Avatar
 
Join Date: May 2012
Posts: 2,291
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 51 Post(s)
Liked: 226
Quote:
Originally Posted by amarshonarbangla View Post

The astronomic differences between synthetic and real world tests is what's confusing me

I see, you may get better feedback by phrasing the question this way. I don't ever run these tests, and I'm not very knowledgeable in regards to what they check

I'm sure that there are users in this forum who can better explain these tools and what could trick them, but they might or might not be using flexraid

Just my $.02

Also, I doubt he's gone and "programmed" a way to trick synthetic benchmarks, but if he did that would make me laugh. Taking a S. Jobs approach to marketing, with "magic" smile.gif
Dark_Slayer is offline  
post #10 of 28 Old 03-04-2013, 08:34 AM
AVS Addicted Member
 
Mfusick's Avatar
 
Join Date: Aug 2002
Location: Western MA
Posts: 21,441
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 62 Post(s)
Liked: 654
I'm on my phone but I'd suggest trying a larger test size. Try 1GB sized test.

It should slowdown when the memory and cache prefetch expire and show you a more realistic score.

-

"Too much is almost enough. Anything in life worth doing is worth overdoing. Moderation is for cowards."
Mfusick is online now  
post #11 of 28 Old 03-04-2013, 09:16 AM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by Mfusick View Post

I'm on my phone but I'd suggest trying a larger test size. Try 1GB sized test.

It should slowdown when the memory and cache prefetch expire and show you a more realistic score.

Check the charts. I tested both 1GB and 2GB.

I have an interesting idea to test FlexRAID some more. Will report back with results.
amarshonarbangla is offline  
post #12 of 28 Old 03-04-2013, 03:04 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Some interesting developments. I noticed significant spikes in RAM usage ONLY when HDD benchmarking tools were ran on the FlexRAID pool. Such spikes weren't present when the benchmarks were ran on a drive outside the pool. That would probably mean FlexRAID offloads the benchmarks to the RAM when it detects such tools. Mfusick might have guessed correctly. These numbers are deceiving. Not sure if it was intentionally coded in by Brahim.

EDIT
Lol, I am seeing a lot more CPU activity also when benchmarks are ran on the FlexRAID pool.
amarshonarbangla is offline  
post #13 of 28 Old 03-04-2013, 07:19 PM
AVS Special Member
 
robnix's Avatar
 
Join Date: Dec 2003
Posts: 1,612
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 253
Quote:
Originally Posted by amarshonarbangla View Post

Some interesting developments. I noticed significant spikes in RAM usage ONLY when HDD benchmarking tools were ran on the FlexRAID pool. Such spikes weren't present when the benchmarks were ran on a drive outside the pool. That would probably mean FlexRAID offloads the benchmarks to the RAM when it detects such tools. Mfusick might have guessed correctly. These numbers are deceiving. Not sure if it was intentionally coded in by Brahim.

EDIT
Lol, I am seeing a lot more CPU activity also when benchmarks are ran on the FlexRAID pool.

Using ATTO - Memory jumps about 150MB and CPU goes up on my system. Flexraid is writing to disk. However - As with you, speeds are simply too fast for spindle based disk.
Using SQLIO - Memory doesn't change, CPU usage goes up. Flexraid is writing to disk. However - speeds are inline with what I would expect from spindle drives.
CIFS Transfer 66GB File - Memory doesn't change, CPU usage goes up. Flexraid is writing to disk, getting around 90MB/sec.

Looky here!
robnix is online now  
post #14 of 28 Old 03-04-2013, 07:35 PM
AVS Addicted Member
 
assassin's Avatar
 
Join Date: Jul 2004
Posts: 12,894
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 205
assassin is offline  
post #15 of 28 Old 03-04-2013, 08:29 PM
AVS Special Member
 
robnix's Avatar
 
Join Date: Dec 2003
Posts: 1,612
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 253
Quote:
Originally Posted by assassin View Post

Ahh. Benchmarks.

Gotta love 'em!

Ain't they great!

Looky here!
robnix is online now  
post #16 of 28 Old 03-04-2013, 08:53 PM
AVS Special Member
 
kapone's Avatar
 
Join Date: Jan 2003
Posts: 4,311
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 9 Post(s)
Liked: 39
- significant spikes in RAM usage ONLY when HDD benchmarking tools were ran on the FlexRAID pool.

+

- Such spikes weren't present when the benchmarks were ran on a drive outside the pool.

=

That would probably mean FlexRAID offloads the benchmarks to the RAM when it detects such tools.

??

You didn't stop to think that maybe that IS how Flexraid works, by caching in memory, whether bench marking or not?
kapone is offline  
post #17 of 28 Old 03-04-2013, 09:16 PM
AVS Special Member
 
robnix's Avatar
 
Join Date: Dec 2003
Posts: 1,612
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 59 Post(s)
Liked: 253
Quote:
Originally Posted by kapone View Post

- significant spikes in RAM usage ONLY when HDD benchmarking tools were ran on the FlexRAID pool.

+

- Such spikes weren't present when the benchmarks were ran on a drive outside the pool.

=

That would probably mean FlexRAID offloads the benchmarks to the RAM when it detects such tools.

??

You didn't stop to think that maybe that IS how Flexraid works, by caching in memory, whether bench marking or not?

The only time I saw an increase in Flexraid RAM usage was when I ran ATTO locally. It jumped up 150MB or so. Running SQLIO locally and a CIFS transfer from my workstation had no impact no Flexraids memory usage.

Looky here!
robnix is online now  
post #18 of 28 Old 03-04-2013, 09:19 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by kapone View Post

- significant spikes in RAM usage ONLY when HDD benchmarking tools were ran on the FlexRAID pool.
You didn't stop to think that maybe that IS how Flexraid works, by caching in memory, whether bench marking or not?

Mhm I did. Assuming FlexRAID does cache to RAM during file transfers, it doesn't seem to help a whole lot. If it did cache to memory, I would assume transferring files would be faster, but it's not. The speeds are about the same when compared to a bare drive, which it should be and not something ludicrous as the benchmarks are suggesting. Caching to memory probably is not the only factor in play here, if it's a factor at all. If it was, real-world tests vs synthetic tests should have produced similar numbers. Maybe I am just worrying too much about something that doesn't matter a whole lot.
amarshonarbangla is offline  
post #19 of 28 Old 03-04-2013, 09:20 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by robnix View Post

The only time I noticed an increase in Flexraid RAM usage was when I ran ATTO locally. It jumped up 150MB or so. Running SQLIO locally and a CIFS transfer from my workstation had no impact no Flexraids memory usage.

Are you sure? I noticed higher memory usage when transferring files over to the pool.
amarshonarbangla is offline  
post #20 of 28 Old 03-13-2013, 09:20 AM
AVS Addicted Member
 
Mfusick's Avatar
 
Join Date: Aug 2002
Location: Western MA
Posts: 21,441
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 62 Post(s)
Liked: 654
I don't think we will find a clear answer unless the dev wants to admit it.

-

"Too much is almost enough. Anything in life worth doing is worth overdoing. Moderation is for cowards."
Mfusick is online now  
post #21 of 28 Old 03-13-2013, 10:04 AM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Brahim won't give us a clear answer. He actually got angry and gave me a warning for trying to force an answer out of him lol.
amarshonarbangla is offline  
post #22 of 28 Old 05-29-2013, 03:03 PM
AVS Addicted Member
 
Mfusick's Avatar
 
Join Date: Aug 2002
Location: Western MA
Posts: 21,441
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 62 Post(s)
Liked: 654
Did you ever retest this ?

-

"Too much is almost enough. Anything in life worth doing is worth overdoing. Moderation is for cowards."
Mfusick is online now  
post #23 of 28 Old 05-29-2013, 05:13 PM
Senior Member
 
Elpee's Avatar
 
Join Date: Dec 2007
Posts: 379
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 17
Quote:
Originally Posted by amarshonarbangla View Post

Brahim won't give us a clear answer. He actually got angry and gave me a warning for trying to force an answer out of him lol.
Wow, I'm safe to go with snapraid. Nothing to lose if I would lose...
Elpee is offline  
post #24 of 28 Old 05-29-2013, 05:49 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by Elpee View Post

Wow, I'm safe to go with snapraid. Nothing to lose if I would lose...

Meh, I understand why he was angry. He was only trying to protect his intellectual property but I was trying to force an answer out of him. Because of this argument with him however, I now have a better understanding of how FlexRAID works.

FlexRAID is still the best snapshot based RAID solution IMO. I wouldn't go with any other product.
amarshonarbangla is offline  
post #25 of 28 Old 05-29-2013, 05:54 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by Mfusick View Post

Did you ever retest this ?

Tested this just now. Same results as before

image206k.jpg
amarshonarbangla is offline  
post #26 of 28 Old 05-31-2013, 09:08 AM
AVS Special Member
 
EricN's Avatar
 
Join Date: May 2002
Posts: 1,167
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 13 Post(s)
Liked: 154
Quote:
Originally Posted by amarshonarbangla View Post

Meh, I understand why he was angry. He was only trying to protect his intellectual property but I was trying to force an answer out of him.

Patents protect IP. Being rude to customers is just being a dick. Your question was legitimate. In my experience, "secret sauce" only comes in two flavors: ******** and snake oil. Actual wizardry (i.e.: 0x5f3759df method) can usually be adequately described without revealing the technical implementation.
EricN is online now  
post #27 of 28 Old 08-19-2013, 02:59 PM
AVS Addicted Member
 
Mfusick's Avatar
 
Join Date: Aug 2002
Location: Western MA
Posts: 21,441
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 62 Post(s)
Liked: 654
It was never determined the reason ? Or was it determined to be a ram as a cache ?

-

"Too much is almost enough. Anything in life worth doing is worth overdoing. Moderation is for cowards."
Mfusick is online now  
post #28 of 28 Old 08-19-2013, 05:17 PM - Thread Starter
AVS Special Member
 
amarshonarbangla's Avatar
 
Join Date: Mar 2011
Posts: 1,239
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 39
Quote:
Originally Posted by Mfusick View Post

It was never determined the reason ? Or was it determined to be a ram as a cache ?

RAM as a cache.
amarshonarbangla is offline  
Reply Home Theater Computers

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off