On Sat, Apr 15, 2006 at 08:31:18AM -0600, David Wilk wrote:
> I've got a test server with OpenBSD-3.8 on it (GENERIC except with
> RAIDFRAME support) and considering the hardware, it does pretty well
> with NFS performance. However, I've noticed that when under heavy NFS
> load, it becomes nearly unresponsive. Shell sessions take 2-6 seconds
> to respond, and top updates stretch out for several seconds as well.
>
> Is this normal behavior (trading latency for throughput)? or is there
> something else going on? Are there common pitfalls with regard to
> OpenBSD kernel tuning and latency?
I don't know about NFS, especially, but it is a common pattern that very
heavy disk activity causes responsiveness to lessen quite a bit.
ISTR reading somewhere that such behaviour is indirectly linked to the
fact that OpenBSD does not support kernel threads. However, I'm not even
sure if the latter is still true...
Joachim
