This may be a disk IO bottle neck on the Linux server. The load average
you showed, is that CPU or Disk IO? Are the drive(s) fragmented,
Raided? Or just a single large drive?
Have you started a monitor on your Ethernet interface(s)?
Generally if the sending system is not CPU or IO bound, the network is
not over taxed, and the receiving server is not CPU or IO bound, you
have some sort of network miss-configuration. These types of problems
can take lots of research and one will need to know the whole picture to
see what is going on and where. They are a pain but fun distraction
from the day to day work.
Director of Information Services
[mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of James Rich
Sent: Friday, August 14, 2009 2:11 PM
To: Midrange Systems Technical Discussion
Subject: RE: NFS performance
On Fri, 14 Aug 2009, Chris Bipes wrote:
The server you are coping to: What is it and how is it configured?
While this process is running is it's disk(s) pegged?
The server being copied to (i.e. the NFS server that physically hosts
NFS exported directory) is Linux 188.8.131.52-smp. top shows the CPU to be
rather idle, but the load looks like this:
load average: 1.62, 1.38, 1.28
That is an usually high load for this machine. Doing NFS copies between
linux machines doesn't generate nearly as much load, so perhaps NFS on
iSeries is different in such a way to cause this high load? In my
opinion, even at this load the machine should still be doing better than
about 1GB per hour.