building a RAID system

Bogdan Costescu bogdan.costescu at
Thu Oct 9 08:24:20 EDT 2003

On Wed, 8 Oct 2003, Mark Hahn wrote:

> what reason do you have to regard 8M as other than a useless
> marketing feature?  I mean, the kenel has a cache that's 100x
> bigger, and a lot faster.

Yes, but the kernel might be dumb at times, like when splitting large 
requests into small pieces to be fed to the block subsystem just to be 
reassembled again before being sent to the disk :-)
Another issue is how this memory is used by the drive firmware. I've seen 
tests that show some Fujitsu SCSI disks (MAN or MAP series, IIRC) perform 
much better than competitors in multi-user situations (lots of different 
files accessed by different users, supposedly scattered on the disk) while 
the competitors were better at streaming media (one big file used by a 
single user, supposedly contiguously placed on disk).

> unless your workload is dominated by tiny, random seeks,

Or your file-system becomes full and thus fragmented. Been there, done 
that! I've had a big storage device changed from ext3 to XFS because ext3 
at about 50% fragmentation was horribly slow; XFS allows live (without
unmounting or mounting "ro") defragmentation.

Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu at IWR.Uni-Heidelberg.De

Beowulf mailing list, Beowulf at
To change your subscription (digest mode or unsubscribe) visit

More information about the Beowulf mailing list