I would like to chat with someone about the tuning document.. I have
a NetPIPE module for PVFS2 now, and I would like to be able to
develop some hooks to allow a user to go run NetPIPE with a bunch of
different tuning options, and change things like FlowBufferSizeBytes
without having to restart all the servers, or at least automate the
process of restarting the servers.
On Oct 19, 2006, at 2:28 PM, Rob Ross wrote:
Sure! I think that the small values really only make sense for
systems where TCP is being used with small kernel buffer sizes.
We're working on a tuning document describing starting points for
many of these parameters on various types of systems.
Rob
Scott Atchley wrote:
On Oct 19, 2006, at 1:12 PM, Sam Lang wrote:
So the 256K chunks you see are from that default flow buffer
size. We only post bmi operations in flow buffer sized chunks,
so at the BMI method layer (IB in this case) on the server, you
see those memory registrations. On the client the buffer passed
in through the system interfaces is used, so I would bet that the
memory registrations don't get chunked into flow buffer sized
regions like they do on the server.
To change the flow buffer size, you can add this to your fs.conf:
FlowBufferSizeBytes 1446864
It needs to be added within the <Filesystem> context.
-sam
This will make a difference on MX as well. If you want to get the
most performance from a 10 Gb/s link, you want bulk transfers of
512 KB or (preferably) more.
Scott
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
