but I'm still shocked to see that it can take close to or over a second to get a response sometimes.
Yes. So maybe counting the % of responses that take > 100 ms(or some other threshold you think is reasonable) might be a more stable way to figure out how bad the outliers are.
Have you checked the disk health? Reallocated sectors can greatly increase seek time.
I've just check the disk health and there isn't a single reallocated sector on any one of the drives in the array. Like I said though, I'm not sure what the "expected" values are for a setup like this, so I don't know if what I'm seeing is the "norm" or something to be worried about; either way the numbers don't look too impressive, but possibly workable. Shame that no-one else could run a comparison test for me using a VMware + ReadyNAS 3200 setup, but I imagine there are few and far between people using this..?
I am busy setting up a ReadyNAS 3200 with VMware as well. Noticed also some odd performance which is starting to make me wonder. Quick run down of my setup:
ESXi 4.1 Hosts
- 2 hosts with each one NIC dedicated to the NAS - No Jumbo Frames - 1000 Mb/s NIC
Switch
- Dedicated VLAN for NAS and ESXi hosts - Cisco Catalyst 3700 series - Jumbo Frame enabled (in case I need to enable it on the NAS and ESXi hosts)
NAS
- ReadyNAS 3200 latest firmware and 12x 2TB X-RAID2 with dual redundancy array with one NFS share - NIC teamed (802.3ad LACP and switch configured)
