Hi there.

I have been doing some raid performance measurements over a range of
configurations, hoping to characterise any bottle necks, and thought I
would share the results.

I only have 7 drives in my test system, and they are on both channels
of a aha3940  U2W scsi controller, so everything applies only
to scsi drives.

Gary Murakami's has some similar performance numbers for IDE at

  http://www.research.att.com/~gjm/linux/ide-i75raid.html

and it was his work that motivated me to do mine.

The tests were to run bonnie and dbench on an ext2 filesystem with all
combinations of:
   raid0 and raid5
   2 to 7 drives
   chunk size from 4k to 256k is exponential steps (is that the right term?)
   patched 2.2.13,  plain 2.4.0-test10, patched (my patches)  2.4.0-test11

In brief, my conclusions were:
  raid0 is much better in 2.4 than 2.2.
    This reflects changes elsewhere in the kernel as the raid0 code is essentially 
unchanged.

  raid5 in 2.4 is much worse than in 2.2, but my patches improve it
     substantially, but there is still work to do.

  2.2 has problems with larger chunk sizes which 2.4 doesn't seem to
    have.


If anyone else has any other favour bench marks that they would like
to run and share the results of, I would be very interested.
I would like to run SPEC/SFS against knfsd over raid, but I fear that
setting that up to automatically reconfigure the server after each run
in the clients is more effort than  have time for at the moment.

My results, complete with pictures, can be found at:

    http://www.cse.unsw.edu.au/~neilb/wiki/?LinuxRaidTest

NeilBrown

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to [EMAIL PROTECTED]

Reply via email to