Egervary Gergely writes:
>  > Another data point - I'm running 2.0_RC4 from October 8, and it's
>  > working just dandy here (XP2000+, 512MB ram, 4 RF mirrors on a pair
>  > of 120G PATA disks).
> 
> Thank you for all your replies. The problem is somewhat strange,
> and I cannot use the proliant box for testing now - it's running
> as a production server w/1.6.2 for 3 days flawlessly.
> (okay, it's just a workaround for now)
> 
> I _really_ want to reproduce this problem on my desktop box, but
> it works good with 2.5-RC5. Hmmm...
> 
> Anyway, if you are running raidframe with RAID level 5, could you please
> test the performance?
> 
> RAID component tests: (Seagate 120GB ATA 7200RPM on ICH5)
> 
> dd if=/dev/rsd0d of=/dev/null bs=512 count=512k shows ~4 MBytes/sec

dd if=/dev/rwd1d of=/dev/null bs=512 count=512k
268435456 bytes transferred in 51.511 secs (5211225 bytes/sec)
dd if=/dev/rwd2d of=/dev/null bs=512 count=512k
268435456 bytes transferred in 43.815 secs (6126565 bytes/sec)

> dd if=/dev/rsd0d of=/dev/null bs=32k count=8k shows ~59 MBytes/sec (very 
> good!)

dd if=/dev/rwd1d of=/dev/null bs=32k count=8k
268435456 bytes transferred in 4.833 secs (55542200 bytes/sec)
dd if=/dev/rwd2d of=/dev/null bs=32k count=8k
268435456 bytes transferred in 7.349 secs (36526800 bytes/sec)

> RAID overall tests: (raid 5 with 3 similar disks as above)
> 
> dd if=/dev/rraid0d of=/dev/null bs=512 count=512k shows ~2.3 MBytes/sec

dd if=/dev/rraid3d of=/dev/null bs=512 count=512k 
268435456 bytes transferred in 60.073 secs (4468487 bytes/sec)

> dd if=/dev/rraid0d of=/dev/null bs=32k count=8k shows ~25 MBytes/sec

dd if=/dev/rraid3d of=/dev/null bs=32k count=8k
268435456 bytes transferred in 7.658 secs (35052945 bytes/sec)

> It should be faster, isn't it?

It would seem to be a little on the slow side...  In my case, the 
disks are different (1/2 of an 80GB and 2 40GBs) so it's a bit hard 
to compare against.

> sectPerSU is set to 128
> (makes no sensible difference with lower or higher value)

What does your disklabel look like?  Does raid0a start at '0' or some 
offset?  A sectPerSU value of 32 or 64 should make a *huge* 
difference (vs. 128) for a 3-disk RAID 5 set, especially for 
writes through the filesystem.

Later...

Greg Oster