This post is more than 5 years old
12 Posts
0
5602
January 7th, 2013 15:00
Help me understand what this Experimental Analysis is trying to tell me
We have several arrays, some SATA and some SAS. The SAS one looks fine when using this tool but the SATA one (above) always says the array is maxed out. This would be a concern except that the IOPS for both read and write climb several times higher that the level where it first shows 100% usage. Likewise the estimated max iops is way below the level of actual iops on many occasions.
I understand that this is generated by an algorithm to be used as a guide and so not perfect but in the case of this array, it's not even close. It never is.
Either a) I'm reading it wrong or b) some profile for the sata array algo is not correct.
Can anyone shed some light on this?
Thanks
No Events found!
Paul_Dwyer
12 Posts
0
January 7th, 2013 20:00
This is ESX, the SAS ones that look okay are also ESX. Replication is not enabled.
"Delayed ACK" has been mentioned to me a couple of times of late, is it enabled by default on the SAN or is it enabled on the switch or in ESX or all of these? Is there a best practice here? I think it is enabled because we see high latency spikes when traffic is close to zero iops implying that it's putting data blocks together, this is not a big deal but I'm not sure if there are other issues or side effects.
end-to-end jumbo frames are being used though
Paul_Dwyer
12 Posts
0
January 7th, 2013 21:00
Thanks, I'll look into these.
Still seems odd that we see these weird stats on the SATA side and not the SAS