Terrible SCSI RAID Performance--Desparate need of HELP

[dariushou12]

Also, why am i getting such crappy ATTO benchmarks on the Adaptec 39160 SCSI controller. Just one device (15K U160 Fujitsu hard drive with only the OS and benchmark programs installed) is installed on the controller. I'm getting a crappy 30MB/sec. It's a basic disk though--can you convert a basic disk to a dynamic disk with the OS on there. OR even from scratch. I'm willing to try anything. There's a thing going around on storagereview.com called the casfilter which is suppose to give you the same perfomance with WindowsXP with a basic disk as with a dynamic disk (due to the cache problem with XP) that is suppose to solve the problem. Not in my case though. I should be getting 60MB/sec atleast with this drive.

Any ideas?

 

[mastertech01]

Well you certainly arent going to find any dual P4 boards, you are correct about that. Dual Athlon would be your least expensive route, but Dual Xeons perform quite well. I know of no single processor boards with 64bit PCI. RIMM4200 is only supported by a couple boards and I see it disappearing completely with the DUAL DDR chipsets coming out.

 

[mastertech01]

You can convert all your disks to dynamic, you just cant change them back. I have seen large improvements in disk speed with dynamic in XP. I saw no difference using dynamic in Win2k. I think you are expecting a little more than you are going to get from a single drive. Hell even my X15s in single drive performance would barely hit 50 and they are the fastest drives on the block last I checked. And fujitsu hasnt had the best reputation lately either. It may even be a firmware issue with the fujitsu drives themselves or bios settings on the drives themselves, hardware configuration, cabling, controller settings, so many things that can affect performance.

 

[Sundog]

Originally posted by: mastertech01
You can convert all your disks to dynamic, you just cant change them back. I have seen large improvements in disk speed with dynamic in XP. I saw no difference using dynamic in Win2k. I think you are expecting a little more than you are going to get from a single drive. Hell even my X15s in single drive performance would barely hit 50 and they are the fastest drives on the block last I checked. And fujitsu hasnt had the best reputation lately either. It may even be a firmware issue with the fujitsu drives themselves or bios settings on the drives themselves, hardware configuration, cabling, controller settings, so many things that can affect performance.

I have had a problem with trying to unmount a drive with XP.:frown:

 

[dariushou12]

The reason i got the 15K Fujitsu Drive over the Cheetah or IBM was becaused they outperformed them in every test on the review on Tomshardware. Doing a new install you can make any disk basic or dynamic with out any regards to what it was before, right? The cabling on these drives are all Adaptec--the good ones--even switched them out with others as i over bought. Tried multiple settings with similar results. I would be happy to hit even 50MB/sec. I can't find any firmware upgrades for these drives--i looked extensively. I will further look into that though.

 

[dariushou12]

when u say bios setting on the drives themselves, you are referring to the scsi bios/ motherboard bios settings, right?

 

[mastertech01]

I dont know what fujitsu has for utilities, but Seagate has utilities you can use to optomize the bios settings within the drive itself. You will have to search the Fujitsu people for that. Some drives are firmware crippled to slow the drive down to make them quieter in operation for example. Quantum was famous for that with thier original 10K. It was so damned noisy you couldnt stand to be in the same room with it...then miraculously they became much quieter after RMA or firmware updated... and much slower.

 

[dariushou12]

by the way if you use or know any thing about ATTO Benchmark scores could u clarify something for me. I know you should use 32MB for the total length field box, but what about transfer size--to get an accurate reading. it goes from .5MB to 4MB. Just wondering here.

 

[kidjan]

This issue has been addressed here a bunch. Try going to their forums; someone has actually made a patch to address the issue (unsupported, of course).

They might also be able to offer a bit more help.

 

[kidjan]

Even more specifically, since the thread is sort of old on SR.com,

check this out for info and a patch

 

[mastertech01]

Originally posted by: dariushou12
by the way if you use or know any thing about ATTO Benchmark scores could u clarify something for me. I know you should use 32MB for the total length field box, but what about transfer size--to get an accurate reading. it goes from .5MB to 4MB. Just wondering here.

Here is a link to one of the benchmarks I did with the ATTO RAID. You can see my settings for the benchmarks in the view. Note that it was with WinXP and dynamic disk. Also this was with XP Pro as the boot ARRAY, not as a storage drive.

 

[Pariah]

I have seen large improvements in disk speed with dynamic in XP. I saw no difference using dynamic in Win2k.

The problem doesn't appear in Win2k until you install SP3.

Hell even my X15s in single drive performance would barely hit 50 and they are the fastest drives on the block last I checked.

X15's are quite dated at this point, there are a slew of drives including ATA drives that break 50MB/s. The newest 15k.3 tops out at 76MB/s.

And fujitsu hasnt had the best reputation lately either.

Fujitsu SCSI drives are very good.

It may even be a firmware issue with the fujitsu drives themselves or bios settings on the drives themselves, hardware configuration, cabling, controller settings, so many things that can affect performance.

None of those are completely crippling the performance of his setup. If he drops WinXP and gets a better RAID card, the problems should go away. Even with 32/33 PCI slots his performance should be over 110MB/s, not just above 60MB/s.

The reason i got the 15K Fujitsu Drive over the Cheetah or IBM was becaused they outperformed them in every test on the review on Tomshardware.

I would use StorageReview for HD benchmarks before THG. Fujitsu drives are very good performers especially in server benchmarks so the drives are not the problem here.

Doing a new install you can make any disk basic or dynamic with out any regards to what it was before, right?

Yes, but dynamic disc isn't a guaranteed fix. I would dual boot Win2k (do not install SP3) and see what kind of performance you get. You'll have a much better idea of what the problems might be that way.