In article <200407150627.i6F6RoDG027016@egsner.cirr.com>,
Stephen M. Jones <smj@cirr.com> wrote:
>This is really interesting to see.  Just a little background, I had installed
>FreeBSD 5.2.1 on two CS20s and had them act as NFS client/servers to
>each other while they continuously wrote large files on each other's
>NFS mounted disks.  The CS20s are basically identical in memory and disk,
>and as you'd expect the numbers for writing on each other's mounts were
>fairly even with numbers between ~2070000 and ~3011000 B/s  .. slow for
>100mbit ethernet, but fair for NFS you might think.
>
>Today, I decided to give NetBSD 2.0x another shot and installed 2.0F on
>one CS20 (the later with the 2.0G kernel) and continued the writing on
>NFS mounts.  Besides the issues I wrote about earlier with the generic
>SMP kernel locking up within an hour then nfsio performance degregation
>and weird locking problems, I've noticed that the NetBSD client writes 
>faster to the FreeBSD server than vice versa.  I should also note that
>numbers I'm seeing from the FreeBSD client writes are a 1/3 or less of
>what they were when it was writing to a FreeBSD server.
>
>Client   Server   Low B/s   High B/s    Avg B/s   Iterations  Generated Zeros
>FreeBSD  NetBSD   1118474     725269     888298           82            200MB
>NetBSD   FreeBSD  1728968    3433563    2421736           82            200MB
>
>Avg is based on the sum of all 82 iterations.  I should also note that 
>although the two machines are using the same network, device type and
>medium, the NetBSD machine has reported 24247 errors and has had a few
>interface timeouts which could have prevented the FreeBSD client from
>writing fast enough .. However, since its the same ethernet device, 
>it obviously can write out well enough to the FreeBSD server.  Again,
>no tweaking, just fresh installs .. so I am assuming that the writes are 32k.

Does "netstat -i" show 0 errors for FreeBSD? Is the driver dignificantly
different?

christos