Throughput issues with RAID array.

[fredweston]

I have a Powervault 200S SCSI storage enclosure with 8 Seagate 50GB 7200 RPM SCSI drives. The drives are Ultra2 (80MB/s) and they are connected to a Compaq DL380 with an Ultra2 RAID controller. The array contains 7 drives in RAID 5 (one is an online spare). Using the quick benchmark in HD Tach, it's only reporting about 30-35 MB/s burst. This seems slow to me since it's in line with what I would expect from a normal IDE drive. I know the drives I have are a bit on the old side, but it seems like an array of 7 should be faster than a single IDE drive.

For comparison, I have a different computer with (3) 18GB 10,000 RPM SCSI Drives and it's bursting at 45 MB/s. That computer is actually quite a bit slower than the first, but the drive and controller setups are virtually identical except for the number of drives and the spindle speed.

 

[PCcashCow]

Which version of HD TECH are you using?

 

[jap0nes]

I have a Powervault 200S SCSI storage enclosure with 8 Seagate 50GB 7200 RPM SCSI drives. The drives are Ultra2 (80MB/s) and they are connected to a Compaq DL380 with an Ultra2 RAID controller. The array contains 7 drives in RAID 5 (one is an online spare). Using the quick benchmark in HD Tach, it's only reporting about 30-35 MB/s burst. This seems slow to me since it's in line with what I would expect from a normal IDE drive. I know the drives I have are a bit on the old side, but it seems like an array of 7 should be faster than a single IDE drive.

For comparison, I have a different computer with (3) 18GB 10,000 RPM SCSI Drives and it's bursting at 45 MB/s. That computer is actually quite a bit slower than the first, but the drive and controller setups are virtually identical except for the number of drives and the spindle speed.

relative to the 10k rpm array, it seems like normal to me... raid 5 is not the fastest setup, considering a 10k rpm raid maxes 45MB/s and the 7200 35MB/s looks acceptable.
Dont forget that a 7200 scsi drive is quite old, and if you're comparing to a moderns 7200 rpm ide drive, they perform close in transfer rate.

 

[fredweston]

Well, not having any actual benchmarks in hand from an IDE drive of the same era (5-6 years ago), I thought the difference in throughput from then to now as far as IDE drives goes shouldn't be that much.

It just seems odd to me that they'd market the Ultra2Wide interface when the actual drives themselves aren't fast enough to saturate regular Ultra2. In other words, it seems like a single drive should be able to get close to 30-35 MB/s on an 80 MB/s interface just by comparison to IDE drives today which get 35-50 MB/s on a 133 MB/s interface. I'm just going by the results from my 300GB WD IDE drive, which isn't anything special and is actually considered pretty slow by today's IDE drive standards.

I realize RAID 5 isn't a speed demon, but any RAID level with striping should be faster than a single drive, so by that logic it seems like if I tested a single drive, I'd get something like 5-10 MB/s? That just seems way too slow, and hence it prompted me to think something wasn't right.

 

[SHv2]

I'd say RAID 5 is decently speedy. On my RAID 5 setup I can pull 100MB/s read and write all the way across the drive.

 

[fredweston]

I'd say RAID 5 is decently speedy. On my RAID 5 setup I can pull 100MB/s read and write all the way across the drive.

How about some specs on what type of drives you have?

 

[SHv2]

How about some specs on what type of drives you have?

I have 4 WDC WD4000YR-01PLB0 hooked up to a FasTrak TX4310.

 

[jap0nes]

oh, and dont forget that the stripe size and block size influence directly on the speed of the array

 

[SHv2]

oh, and dont forget that the stripe size and block size influence directly on the speed of the array

Do you happen to know of a site that shows comparisns between these kinds of things?

 

[jap0nes]

oh, and dont forget that the stripe size and block size influence directly on the speed of the array

Do you happen to know of a site that shows comparisns between these kinds of things?

http://www.anandtech.com/storage/showdoc.aspx?i=1491&p=5

http://www.somacon.com/p426.php

http://www.hardware-one.com/reviews.asp?aid=148&page=4

http://www.tldp.org/HOWTO/Software-RAID-HOWTO-9.html#ss9.3

the last one was run on a linux software raid, but i guess the results are proportional to a hardware raid. Note that raid 5 is the slowest setup

 

[SHv2]

http://www.anandtech.com/storage/showdoc.aspx?i=1491&p=5

http://www.somacon.com/p426.php

http://www.hardware-one.com/reviews.asp?aid=148&page=4

http://www.tldp.org/HOWTO/Software-RAID-HOWTO-9.html#ss9.3

the last one was run on a linux software raid, but i guess the results are proportional to a hardware raid. Note that raid 5 is the slowest setup

Thank!