Poor SATA / RAID performance (2.6.11 and promise SATAII150 TX4)
Greetings,
I'm attempting to benchmark software RAID5 on a system with:
- Promise SATAII150 TX4 card
- 4 Segate ST3300831AS drives
- custom built kernel 2.6.11 (to get driver for promise SATAIITX4)
- FC3 install
- EPIA M10000 mainboard, 256MB memory
The tools I'm familiar with for benchmarking a PATA based RAID system are:
- hdparm
- dd
Here are some interesting stats from my system:
[root@epiam10k ~]# more /proc/mdstat Personalities : [raid5]
md0 : active raid5 sdd1[3] sdc1[2] sdb1[1] sda1[0]
879100608 blocks level 5, 4k chunk, algorithm 2 [4/4] [UUUU]
unused devices: <none>
[root@epiam10k ~]# hdparm -t /dev/sda
/dev/sda:
Timing buffered disk reads: 116 MB in 3.02 seconds = 38.45 MB/secHDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
[root@epiam10k ~]# hdparm -t /dev/sda1
/dev/sda1:
Timing buffered disk reads: 104 MB in 3.05 seconds = 34.10 MB/secHDIO_DRIVE_CMD(null) (wait for flush complete) failed: Inappropriate ioctl for device
[root@epiam10k ~]# hdparm -t /dev/md0
/dev/md0:
Timing buffered disk reads: 72 MB in 3.03 seconds = 23.79 MB/sec
[root@epiam10k ~]# vmstat
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa 1 0 0 3628 8984 142704 0 0 423 23 1023 112 30 3 61 6
Observations:
- the performance of a raw SATA device (/dev/sda in the above example) seems
low when I compare it to a PATA drive from a previous system (which would get about 45MB/sec)
- the performance of the RAID5 array (/dev/md0) seems very low - I expect
quite an increase over a single device due to striping
- the number of interrupts per second (1023) seems very high
Questions:
- is hdparm the right tool for looking at SATA devices?
- is the error regarding the ioctl an issue?
- how do I set/get the DMA modes for the SATA controller and SATA drives?
- why would the interrupts per sec reported from vmstat be so high?
Thanks for any suggesitons. Please 'cc' me in any replies to the list.
Tim Harvey
Quoted 9 lines · Tim Harvey
Greetings,
I'm attempting to benchmark software RAID5 on a system with:
- Promise SATAII150 TX4 card
- 4 Segate ST3300831AS drives
- custom built kernel 2.6.11 (to get driver for promise SATAIITX4)
- FC3 install
- EPIA M10000 mainboard, 256MB memory
Quoted 6 lines
The tools I'm familiar with for benchmarking a PATA based RAID system are:
- hdparm
- ddHere are some interesting stats from my system:
Here goes my setup/results, FWIW...
Hard: Dual PIII@9333
2 Promise FastTrak S150 TX4 (rev 02)
6x Maxtor 7Y250M0, 250Gb, 3 on each card.
Soft:
Quoted 6 lines
[root@epiam10k ~]# more
/proc/mdstatPersonalities : [raid5]
md0 : active raid5 sdd1[3] sdc1[2] sdb1[1] sda1[0]879100608 blocks level 5, 4k chunk, algorithm 2 [4/4] [UUUU]unused devices: <none>
nada:~# cat /proc/mdstat
Personalities : [linear] [raid0] [raid1] [raid5] [multipath] [raid6] [raid10]
md0 : active raid5 sdf1[5] sde1[4] sdd1[3] sdc1[2] sdb1[1] sda1[0]
1225557760 blocks level 5, 256k chunk, algorithm 2 [6/6] [UUUUUU]unused devices: <none>
Quoted 8 lines
[root@epiam10k ~]# hdparm -t
/dev/sda
/dev/sda:
Timing buffered disk reads: 116 MB in 3.02 seconds = 38.45 MB/secHDIO_DRIVE_CMD(null)(wait for flush complete) failed: Inappropriate ioctl for device
[root@epiam10k ~]# hdparm -t/dev/sda1
Quoted 4 lines
/dev/sda1:
Timing buffered disk reads: 104 MB in 3.05 seconds = 34.10 MB/secHDIO_DRIVE_CMD(null)(wait for flush complete) failed: Inappropriate ioctl for device
nada:~# hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 900 MB in 2.00 seconds = 449.84 MB/sec
Timing buffered disk reads: 172 MB in 3.03 seconds = 56.70 MB/sec
Quoted 5 lines
[root@epiam10k ~]# hdparm -t
/dev/md0
/dev/md0:
Timing buffered disk reads: 72 MB in 3.03 seconds = 23.79 MB/sec
nada:~# hdparm -tT /dev/md0
/dev/md0:
Timing cached reads: 912 MB in 2.01 seconds = 454.48 MB/sec
Timing buffered disk reads: 178 MB in 3.04 seconds = 58.60 MB/sec
Quoted 5 lines
[root@epiam10k ~]# vmstat
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----r b swpd free buff cache si so bi bo in cs us sy id wa 1 0 0 3628 8984 142704 0 0 423 23 1023 112 30 3 61 6
Quoted 8 lines
Observations:
- the performance of a raw SATA device (/dev/sdain the above example) seems
low when I compare it to a PATA drive from a previous system (which would get about 45MB/sec)
- the performance of the RAID5 array (/dev/md0) seems very low - I expect
quite an increase over a single device due to striping
- the number of interrupts per second (1023) seems very high
It is not marvelous in my case, but slightly better than yours...
--
J.A. Magallon <jamagallon()able!es> \ Software is like sex:
werewolf!able!es \ It's better when it's free
Mandrakelinux release 10.2 (Cooker) for i586
Linux 2.6.11-jam10 (gcc 3.4.3 (Mandrakelinux 10.2 3.4.3-7mdk)) #1
Quoted 1 line · Tim Harvey
- the number of interrupts per second (1023) seems very high
Why? You have 1000 timer interrupts every second, plus 23 from other sources.
Lee