I was never really happy with the performance I was getting on the Readynas 4200 about 95 MBs read. It was OK but not great for having multiple nics. So I finally did some optimization this weekend.
The trick was to turn off teaming on the Readynas and have each NIC on a separate Vlan. Then setup ESXi for MPIO round robin in the GUI with 4 NICs on separate Vlans. You would think this is enough but it is not. I was still getting 95MBs performance but instead of seeing 1 nic 100% utilized I saw 4 NIC's 25% utilized. This had me stumped till I found an article about MPIO round robin IOPS and how by default it is set to 1000. It needs to be set to 1. So this ESXi 5 command must be run on each host from SSH.
for i in `ls /vmfs/devices/disks/ | grep naa.600` ; do esxcli storage nmp psp roundrobin deviceconfig set -d $i --iops 1 --type iops;done
Once I did that my read performance went from 95MBs to 400MBs with 4 NICs using the vmware OpenPerformanceTest.icf on IoMeter.
Now I am thrilled with the performance of the Readynas 4200 and feel comfortable virtualizing my Exchange server.
A detailed step by step guide has been posted here https://sites.google.com/site/abraindump/readynas-vmware-iscsi-performance
The iops change increased the throughput I was getting on a Pro 6 from 135MB/s to 228MB/s.
I did a little a further testing using different iops and found that a setting of 4 gave me the best performance... at least for the "Open Performance Test, Max Throughput - 100% Read" test. Not a huge increase but measurable and consistent.
It wasn't very much at all... less than 1MB/s so nothing to get excited about. I originally thought it was just an insignificant variation but 4 was consistently a bit higher than the rest across a number of runs so it must just work slightly better in my setup.