I'm currently working on optimizing our P4500 60TB 5 Node Lefthand SAN, Flow control is enabled as well and each of the 5 nodes are running 2 802.3 bonded 1Gb Nics, the SAN is running version 9.5.
Most of my hosts connecting to the SAN are running single 1Gb nics with the updated HP 9.5 DSM Installed. Currently however, I have a test machine up and running connected to the HP SAN via an MPIO setup with 2 1Gb NIC's connected to the SAN with the method described in a thread not too much further down.
I can confirm that both NIC's are indeed working on the server, and the 802.3 bonding is working on the SAN and configured properly on the switch. However, it sems like we're seeing some pretty bad write performance (reads look great though).
My main question I wanted to ask, is there a standardized IOMeter test I can run on the HP SAN to compare results with others? Currently I have the Outstanding IO Options to 120 (Since with have 12 disks per node * 5 nodes then times 2) just running all of the standard IO tests for 2 minutes each test. The tests are 104800 (1GB?) In size.
It does appear on running all 20 of the standard IOmeter tests that I am indeed breaking the limitations of 1Gb (I've attached a screenshot). And I'm seeing bursts of 120% total network utilization between both NIC's on the server end.
When you say the tests are 104800 in size, are you saying your transfer request size is that big? A 64KB or 8KB would be more realistic depending on what you're trying to simulate.
