I have a CU with the very same issue. I'm adding network alias here as
aggregation is more as a network topic.
Thanks,
Jack
Tim Wissman - SSE wrote:
Folks - i have a client with a 7310 (note, not clustered). their
network switch is setup with LACP as
are 3 of the on board NIC interfaces (nge1,nge2,nge3) and nge0 is just
off on it's own. The data link
configuration is:
using Jumbo frames
Policy L4
Mode active
Timer short
LACP aggregation
The customer is doing the following to gage throughput:
time /opt/sfw/bin/dd bs=1M if=/dev/zero
of=/net/nas1/export/stuff/testing/filename count=10000
(where nas1 resolves to the ip address of the aggregate)
when doing this they see about 97 to 100 MB/s however, when they run
this across multiple systems
at the same time (4 systems have been the max they have tried) the
throughput drops to about 25 to 29 MB/s.
This is what i would expect on a single link, but not on the aggregate.
While they were running these tests, we were also watching the network
status. the traffic was traversing one
link of the aggregate, and the other 2 sat vertually idle. The network
switch is a Cisco box and i believe the
LACP is set to auto; should i have it changed to L4 in order to force
the L$ policy?
NOTE: there was one other test that was run to make sure we had things
configured properly. there was the
same test run as above on two systems, however, one was writing to the
nge0 interface and the other was writing
to the aggregate. the results were what i would expect, 97 to 100 MB/s
on each connection.
Any other thoughts or suggestions would be helpful.
thanks for your time.
tim
_______________________________________________
networking-discuss mailing list
[email protected]
