I'm doing a POC in a costumer, and that's my first time collecting metrics from 3PAR Storage Systems. The costumer usually collects volume metrics from Vcenter, in order to check the latency from datastores and those metrics comes with a satisfatory results as 10, 15 or 20ms. I noticed that there aren't any metrics for 3PAR Storage Pools Latency, controllers and physical disks as well. I'm trying to do a baseline using average read and write IO response time as a substitute for "system service" which is the preferencial metric used to collect "latency" from 3PAR Arrays on 3PAR analisys tool. My question is: On EVA, reasonable latency and response times are accepted between 5~10ms and 15~20ms respectively, but in SE Performance Manager, 3PAR the average IO response time is always bringing me up around 100 ms as a baseline. So, I'd like to understand how SE collects this values cause if there's no latency metrics for 3PAR on SE, how to use response time as a trustable metric in order to create my baselines?
Attached, I'm sending 2 screenshots. One contains 3PAR response time from SE, and on onther contains, datastores latency. Of course they're different metrics, but I'm wondering if I can replace this kind of view from Vcenter by some metric on SE.
I think you are seeing a known issue... the I/O response times formula which were introduced in SE 9.6.0 for 3PAR are inaccurate. I suggest you would need to update to SE 9.6.1 and log a support case for this issue and request a fix.
Wiley Thrasher Storage Management Expert HP Storage TCE and Quality Tiger Team WW L3 Pre-sales Support Hewlett-Packard Company