slow RAID-5 on FreeBSD 5.4

[Tamouh H.]

Hello,

I've just ran Bonnie on a FreeBSD 5.4 machine with Adaptec 2230SLP RAID-5 SCSI Card and Seagate 146GB 10K RPM U320 drives. 

The results were very surprising. The RAID-5 is EXTREMELY slow, and I can't figure out why! 

The RAID controller has two volumes: RAID-5 and a lone drive.

This is Bonnie results from the lone drive on the Adaptec card:

impala# ./Bonnie -d /tmp
File '/tmp/Bonnie.30883', size: 104857600
Writing with putc()...done
Rewriting...done
Writing intelligently...done
Reading with getc()...done
Reading intelligently...done
Seeker 1...Seeker 2...Seeker 3...start 'em...done...done...done...
              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
          100 27168 28.2 31668 13.9 36530 10.8 83540 92.1 495217 98.5 26086.8 182.8

Here are the results doing the same test on the RAID-5:

impala# ./Bonnie -d /usr
File '/usr/Bonnie.30972', size: 104857600
Writing with putc()...done
Rewriting...done
Writing intelligently...done
Reading with getc()...done
Reading intelligently...done
Seeker 1...Seeker 2...Seeker 3...start 'em...done...done...done...
              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
          100  7478  7.7 10727  4.3  7250  3.2 101910 99.6 1464070 99.2 24067.2 180.6


Notice the 7MB write speed compared to 30MB. The read speeds are still very slow considering these are SCSI drives (I've done another test on S-ATA drive and it had the same Read speed, but much slower write).

I believe the drives are working fine, I don't suspect problems with them. However, smartmontools doesn't support Adaptec RAID.

Anything suggested to imrpove the RAID-5 performance ? It is surprising both tests were ran against the same card.

Thanks,

Tamouh Hakmi