I have a Promise VTrak E610fD SAN at home for my 'sandbox' environment. Prior to it, I had an IBM DS3400 SAN, but had horrible performance with it likely due to non-IBM Drives. Overall speed was so-so (about 45MB/s) and based on normal usage I thought it was going to be ok, but turned out when everything was running on it, even setting up another VM (which took easily double the time it took on internal storage) caused everything to slow down. I took that back and now have the Promise SAN, and I'm trying to get a good bead on its overall performance and see if there's anything I need to reconfigure before I put it into play. I've run speed tests using ATTO Disk Benchmark, as well as IOZone and HD Tune. Initially, right after the array finished building, I got good raw speeds, mostly up over 200MB/s with ATTO, but now it's dropped down to around 110MB/s for some reason.
After doing some reading, it appears raw speed is less important than IOPS, so I ran IOMeter on several identical VMs. One on Internal storage, one on the SAN and one on a VM in my work environment on an EqualLogic PS4000e SAN. Both environments are running ESXi 4.1 with the latest patches, but my environment is the free hypervisor and at work I have vSphere Essentials Plus, hence the two VMs on my server due to the pain it is to move VMs between stores with the free version.
After running IOMeter on all three guests with identical settings, the results I got are:
Promise SAN (four 7200RPM 1TB SATA-II in RAID10):
Total IOPs: 3162.99
Total MB/s: 40.60
Average I/O Response time: .6303
Max I/O response time: 264.9995
Internal Storage (three 7200RPM 1TB SATA-II in RAID5):
Total IOPs: 1206.55
Total MB/s: 15.47
Average I/O Response time: 1.6555
Max I/O response time: 218.5185
PS4000e (14 +2 HS 7200RPM 500GB SATA-II in RAID50):
Total IOPs: 2187.22
Total MB/s: 28.03
Average I/O Response time: .9136
Max I/O response time: 105.6791
I ran the tests several times and each time the results were similar.
Based on these results, is it safe for me to assume that this Promise SAN will perform equal to (or based on the numbers, better than) the PS4000e, even though the raw speed tests from within a VM are not near the 4gb FC Link's capabilities (should max out around 256MB/s, and the drives are 3gb which should max out around 384MB/s in theory), or is there some fundamental flaw I'm overlooking that makes these results worthless?
If the ATTO benchmark and IOMeter aren't really good for testing raw speed & IOPs from within a guest, are there any good tests I can run that will give me good, accurate results of the SAN's capabilities?
Thanks,
Dave