How does one interpret Ready Time summations with custom time frames?

[kwg66]

I've been looking at stats for years and thought I was pretty good at understanding them, but I had never thought about, or needed to, look at Ready time historically into a specific time frame with vCenter perf charts. 

Lets say I select a time frame between 12noon and 1pm for the prior day?   Would I take the summation values and /200 because the interval is only 1 hour?  Not sure how the stats are presented when you select an historical time slot..

Any ideas?

[daphnissov]

Ready times in vCenter performance charts are based on 20 second time slices which are then rolled up on an average basis after the most current time interval. They're much more difficult to interpret historically with just vCenter alone unless you have something like vROps in place.

[kwg66]

Yes,  this I know.   However, the fact that you can drill into the history and select a time period at all is indication that VMware should be providing some guidance on how to properly interpret this.   If I select a 1 hour time frame from 1 week ago, how would that roll up when it comes to the calculations needed to understand the values presented?  

[daphnissov]

This is probably in the documentation but is one reason why vCenter performance charts are not appropriate for historical analysis purposes.

[ThompsG · Accepted solution]

Hi there,

Its based on the Statistics Interval that your custom range falls into.

So using your example and taking 1 hour from a week ago (and assuming it is less than 7 days from the current time) then your statistics points are every 30 minutes and are based on being summed up from the previous Statistics Intervals. So you will be presented with a number of data points (at 30 minute intervals) between this period - interestingly you seem to always lose the first data point if you make the custom interval exact, i.e. 12pm based on your scenario, so need to make the start time 11:59AM to have all the points over the 1 hour period.

Now as to interpretting this, I'll use an export from one of our VMs for the time period you mention:

Assuming you want to try and look at something like %RDY for that period then the formula would be: ((Statistics Interval Duration * Seconds in a Minute * ms in a Second ) / Ready ) * 100

Putting some numbers into this would give the following (using the Interval duration for the period shown above which is 30 minutes): I'll use the Ready Average as my data point so that would give (( 30 * 60 * 100) / 7467) * 100) or 4.15% ReadyTime averaged over that hour.

Now remember this is heavily normalised and as daphnissov mentioned this might not be very helpful depending on what you are looking at or trying to chase.

Does that help? Have I made things more confusing?