Re: NetBSD disk I/O

[Andy Ball <ball%grex.org@localhost>]

Hello John,

  JDB> A literal pile of 2.5-inch SATA disks, you say?

It's true!

  JDB> If they were pulls from a hardware RAID box or some
     > such, the disks' firmware may disable write caches by
     > default.

They are pulls but were used individually, not as part of an
array.

  JDB> ...dismal until I enabled the caches with
     > $ dkctl /dev/rwdNd setcache rw

# mount /dev/wd1e /mnt/disk0
# cd /mnt/disk0
# dd if=/dev/zero of=TEST bs=32768 count=32768
32768+0 records in
32768+0 records out
1073741824 bytes transferred in 230.022 secs (4667996 bytes/sec)
# rm TEST
# dkctl /dev/rwd1d setcache rw
# dd if=/dev/zero of=TEST bs=32768 count=32768
32768+0 records in
32768+0 records out
1073741824 bytes transferred in 235.634 secs (4556820 bytes/sec)
#

It's a promising theory. I'm going to find another operating
system to try the same sort of test on, to see whether that
makes a difference.

73,
-Andy Ball