On Wed, May 30, 2007 at 12:49:57PM +0100, David Brownlee wrote:
> 	Its certainly worth benchmarking. If you have a chunk of memory
> 	in the box you might also want to try a tmpfs


I started a bonnie++ run, but afer 1.5 days killed it ;-)

Here are simpler results:

On nfs: /usr/bin/time  dd if=/dev/zero of=./test bs=1m count=64
64+0 records in
64+0 records out
67108864 bytes transferred in 123.054 secs (545361 bytes/sec)
      123.67 real         0.03 user        70.36 sys

On local scsi disk: /usr/bin/time dd if=/dev/zero of=/tmp/test bs=1m count=64
64+0 records in
64+0 records out
67108864 bytes transferred in 98.470 secs (681515 bytes/sec)
       98.86 real         0.04 user        88.95 sys

This is on a 

Apple Macintosh Centris 660AV  (68040)
total memory = 45056 KB
avail memory = 40672 KB

with /tmp on 

sd1 at scsibus0 target 4 lun 0: <IBM, DNES-309170W, SA30> disk fixed
sd1: 8748 MB, 11474 cyl, 5 head, 312 sec, 512 bytes/sect x 17916240 sectors
sd1: async, 8-bit transfers, tagged queueing

and nfs via mc0 at obio0.


Martin