If a lot of random I/O is being flushed to disk, or just multiple writers /
scattered data, IDE disk performance seems to be poor due to (disk) time
spent seeking when writing back dirty buffers.
- If bdflush were to sort the buffer chain when woken, before flushing some
buffers to disk, the gains would be obvious. On a SCSI sub-system, this
would probably give little difference, but on IDE disks, flushing takes much
longer than the transfer rate of the disk x quantity of data to flush.
IMHO, IO bound operations would benefit significantly (eg gunzip'ing). I
expect the additional processor overhead not to be much, since bdflush isn't
woken all that often (every >5 secs under normal load?).
If it were a kernel compile-time option, systems with SCSI disk sub-systems
could disable this feature.
Any feedback?
__________________________
Daniel J Blueman - daniel.j.blueman@stud.umist.ac.uk
Undergraduate - BSc Computing Science
UMIST - Manchester
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/