Hi folks. We've got a particular VM hosting a workload that *very* sensitive to latency/contention and I'm setting up an alarm based on CPU Ready Time. I understand the mechanics of creating the alarm and the alert actions, I get how to convert from CPU Ready in ms to percentages, etc. Unfortunately I'm hung up a bit on what values are appropriate for the Warning/Alert value(s) and the Condition Length. To wit:
Let's say I want an alert (Warning) if CPU Ready hits 3%. Due to the workload profile I know that I need a short Condition Length...let's say 30 seconds. I know that in the Performance tab the way I'd calculate this is to divide the CPU Ready summation value in ms by 20000 (if real-time view) and multiple by 100. So 600/20000*100=3%
But is that the way things work in the Alarms area? If my Condition Length is 30 sec (30000 ms) and I want to alarm at 3% then I'd want a Warning value of 900ms, right? If not, what's the appropriate value and how do I get there?
I've Googled like mad trying to find the answer to this (and searched the forums here and RTFM'd) but haven't found a clear...or really any...explanation of this.
Thanks!
***EDIT***
I guess another way to state this is that I need to know what the collection interval is for Alarms. Are we talking about the 20 sec "Real-Time" interval for collection, subsequently evaluated to be true for more/less than the Condition Length time? If that's the case then I'd want to set my Warning value to 600ms (3% of 20000ms) and I'd know that the value would need to stay there for 2 complete collection intervals (40 sec total) in order to hit the Alarm Condition Length of 30 sec.