Jump to content

Change Mode

Need your RAID scores in ATTO


Recommended Posts

Yeah...the info is just ENDLESS on how people setup there raid array...ENDLESS... :thud:

 

Wow, really? :blink: . . . Anyone care to throw a few pointers in this direction? I don't really know what else could be changed besides strip size, cluster size (I think that's what it's called when you're formating -- Windows does 4k by default I believe), turning Restore off, turning NCQ on, changing the page file (which BF2 really didn't like), and setting up Windows on a separate partition. . . .

 

Is there anything else?

Link to comment
Share on other sites

  • Replies 147
  • Created
  • Last Reply

Top Posters In This Topic

Ok, there is a reason for all these tests and why I asked for this particular setting for ATTO. First, I needed as many different configurations as possible to give us a wide spectrum to analyze.

 

The test is showing a trend, the trend seems to be that there are sometimes a bottle neck happening on some HDDs or sometimes, the controller. For instance, if running 4 SATA1 HDDs on a controller capable of 3Gb, it seems to bottle neck at the HDD. Second is when running 4 fast SATA1 HDDs on a controller capable of 150 or 133, it will bottle neck at the bus.

 

I am looking to find as good a setting for each particular set up as possible, this way we will have something to refer to when advising others or when we are upgrading later.

 

The SATA2 HDDs using 3Gb controllers are definitely the kings here and so they should be. When using over lapped I/O, it stresses the flow of all the hardware involved and these new controllers should be capable of running up to 8 SATA2s in RAID 0 before reaching tis bottle neck on either the HDDs or controller. The burst will be increased by the later release of SATA2 32Mb cache HDDs.

 

Please keep the posts coming as this is has been very informative. :)

 

Final analogy to come soon.

 

X_C

Link to comment
Share on other sites

So does this mean that it’s better to use drives with a wider bandwidth (SATA II versus SATA I) rather than drives with a higher RPM (10,000 versus 72,000)? I was under the impression that RPM has more impact on daily performance because the interface only affects how fast the cache is "dumped." . . .

 

For example, with my setup, I have a cache of 8MB x 4 hard drives = 32MB. Since these drives are SATA I (150MB/s), then the time it takes to dump the cache is

 

32MB divided by 150MB/s = 0.21s.

 

If these drives were SATA II (300MB/s), then the time it would take is

 

32MB divided by 150MB/s = 0.11s.

 

So in daily operation, there would only be a difference of 1/10 of a second which probably would not be noticeable. At that point, the difference in RPM between 7200 and 10,000 would kick in. The Raptors should be faster by a factor of 10,000 / 7200 = 1.38 per drive. If there are four Raptors, then 1.38 x 4 = 5.5. . . . That means that the Raptors would be 5.5 times faster than the 7200 RPM drives. In other words, if it takes 30min. to transfer a file with four 7200 drives in RAID 0, it would take about 5.5min. with four Raptors in RAID 0. Or, for a more practical example, if it takes 5 min. to load a game with 7200 drives, then it would take less than 1 minute with the Raptors. . . . Honestly, I don't know if this analysis is correct, so if anyone sees a mistake please correct me. . .

 

There's one more point that I thought is relevant. . . . If a file is 32MB or less, then couldn't it easily sit in the RAM? So the performance for small files would be affected mainly by the speed of the RAM and not the hard drives, right? . . . Then that would mean that during daily operation, small files (like internet files) would mainly be affected by the RAM speed and large files (like game maps) would mainly be affected by the RPM of the hard drives. The cache size (whether 8MB or 16MB per drive) and the interface (whether SATA I or SATA II) shouldnt play that big of a role, right?

 

As far as the ATTO test, I noticed that we''re only transferring 1MB with "small files" (16kb to 1024kb). Then I assume we're only using the hard drive cache. We're not drawing data directly off the spinning platter so the RPM wouldn't matter at all. I would expect that the test would only be affected by three things . . .

 

1. The interface speed (whether SATA I or SATA II);

2. The number of drives in RAID 0 (since the files can be transferred to more drives simultaneously); and

3. The stripe size (which would affect the comparison between the individual scores for 16kb, 32kb, 64kb, . . . .up to 1024kb. Although, I don't know what stripe size would be best.)

 

Wow, that was a super long post. Sorry. :blushing: I guess I got carried away thinking about this stuff. . . . If anyone can comment on whether any of this stuff is actually right, it would be much appreciated.

Edited by Revivalist
Link to comment
Share on other sites

I think the direct I/O part of ATTO means it uses the disk, not the cache.

If it was the cache it would probably be much faster!

 

Wow, then I'm really surprised. . . . So is the comparison I described between 10,000 and 7200 RPM drives totally wrong?

 

Also, this raises another question . . . should I expect a maximum theoretical performance of 150MB/s x 4 = 600MB/s (based on the interface) or 72MB/s x 4 = 288MB/s (based on the sustained speed that they're rated at)?

Edited by Revivalist
Link to comment
Share on other sites

Yes, pretty much. Also if you have four 10,000rpms it would only be 1.38 times faster than the four 7200s, not 5.5 times.

 

The maximum performance would be around the bus speed, but the actual performance (as we know it) would be the sustained speed

Link to comment
Share on other sites

Yes, pretty much. Also if you have four 10,000rpms it would only be 1.38 times faster than the four 7200s, not 5.5 times.

Hmmm. . . . So having four Raptors wouldn't give a benefit over having one? In other words, if one Raptor is 1.38 times faster than one 7200 drive, then wouldn't four Raptors be 1.38 x 4 = 5.5 times faster than four 7200 drives?

 

The maximum performance would be around the bus speed, but the actual performance (as we know it) would be the sustained speed

 

Ok, so I should expect a theoretical maximum of 288MB/s? But it looks like my test is only getting a little more than half of that. :huh:
Link to comment
Share on other sites

To you first question, NO! Sorry, I just had to shatter that illusion....

Take one 7200 drive to be 50 mb/s ( a little on the slow side, but this is just for explaining) 4 x 50 is 200mb/s

 

With four raptors it's about 4 x 72 which is 288mb/s

 

So it's around 1.44 times faster.

 

As I said theoretical MAXIMUM of the bus speed.

Actual speed (sustained) would be 288 (if you were VERY lucky)

Link to comment
Share on other sites

To you first question, NO! Sorry, I just had to shatter that illusion....

Take one 7200 drive to be 50 mb/s ( a little on the slow side, but this is just for explaining) 4 x 50 is 200mb/s

 

With four raptors it's about 4 x 72 which is 288mb/s

 

So it's around 1.44 times faster.

Ahh, ok, that makes sense. . . .

 

As I said theoretical MAXIMUM of the bus speed.

Actual speed (sustained) would be 288 (if you were VERY lucky)

 

I guess I'm not very lucky. :blink:
Link to comment
Share on other sites

Here is a little info on cache, I thought those that want to read up on it would be interested:

 

To give you the big picture of it all, here's a list of a normal caching system:

 

L1 cache - Memory accesses at full microprocessor speed (10 nanoseconds)

L2 cache - Memory access of type SRAM (around 20 to 30 nanoseconds)

Main memory - Memory access of type RAM (around 60 nanoseconds)

Hard disk - Mechanical, slow (around 12 milliseconds)

Internet - Incredibly slow (between .5 second and 3 days, unlimited size)

 

Modern hard disks come with fast memory, around 512 kilobytes, hardwired to the hard disk. The computer doesn't directly use this memory -- the hard-disk controller does. For the computer, these memory chips are the disk itself. When the computer asks for data from the hard disk, the hard-disk controller checks into this memory before moving the mechanical parts of the hard disk (which is very slow compared to memory). If it finds the data that the computer asked for in the cache, it will return the data stored in the cache without actually accessing data on the disk itself, saving a lot of time.

Obviously writen a while back but it still holds true.

 

X_C

Link to comment
Share on other sites

No offence taken, bigchrome. I already know I'm not as wise about computer stuff as many others are . . . . But that's why I'm here to learn and to also share whatever I do know. :)

 

I agree that Raptors are priced more than what they should be. I was under the impression that they're pretty much the best peformers out there. I found most of mine around $130 so I went for it. Hopefully I can make some of that back when I sell them some time down the line. Until then, I'm still trying to figure out whether 7200 with SATA II is better than 10,000 with SATA I. . . .

 

Thanks for the info, X_C. Yeah, I can tell it's pretty old (0.5MB cache?!) . . . But the operation is still essentially the same. . . By the way, have you been following up on our discussion? Do you have any comment about the 7200 SATA II versus 10,000 SATA I comparison?

Link to comment
Share on other sites

Sorry, I'm at work but I will comment:

 

7200 and 10000 RPM decreases access time

 

Cache is burst and helps with sustained data flow

 

SATA2 and SATA1 is connection speed

 

So far the fastest drives are the SATA2 HDDs, that was until the WD Raptor SATA2, only problem I forsee for them is the bottle neck we are speaking about, if cache can not keep up, it may cause it to slow down.

 

Still waiting to see, hope Sho gets two more before this thread is over. :pullhair: Then we'll know for sure.

 

X_C

Edited by X_C
Link to comment
Share on other sites

So far the fastest drives are the SATA2 HDDs, that was until the WD Raptor SATA2, only problem I forsee for them is the bottle neck we are speaking about, if cache can not keep up, it may cause it to slow down.

 

Still waiting to see, hope Sho gets two more before this thread is over. :pullhair: Then we'll know for sure.

 

X_C

 

Thanks for the comment. Take your time, there's no rush. You can ellaborate later if you care to . . .

 

So are you basically saying that having SATA II on a 7200 drive would more than make up for the slower RPM compared to 10,000 RPM on a SATA I drive? Would this be true for everything, even longer file transfers (like say, anything above 50MB)?

 

Or is it more accurate to say that SATA II with 7200 RPM will only show better performance for small files (like for internet browsing, launching small programs, etc.) and SATA I with 10,000 RPM will do better with larger files (like loading a game, transfering music or video, etc.)

Link to comment
Share on other sites

Well I just got another atto score, I'd posted early with a screen shot and thought my speeds were slow for sata-2, did some research and found out I had to download the feature tool from hitachi to enable sata-2, so I did that and wow, this was great before, but is blazing now. The test is on 4 x hitachi sata-2 80GB, 64/4 on a 12gb boot partion, with nvidia 6.53 drivers, xp with sp2, and restore off. Hope this helps x_c, and thanks for the great work.

 

Posted Image

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share


×
×
  • Create New...