Calculating the change between data points

[JasonGillis]

Hi all,

Now that I've got SQL server metrics feeding into vCOps, I need to work with them to be able to make sense of them.  A lot of the metrics are cumulative since the server start (i.e. batch requests/sec is not a per second metric, it's cumulative), so I need to be able to calculate the change between data points.

I've been trying to figure out how to create a super metric for this, but haven't been able to figure out how.  For example, for a given metric, I want to create a super metric that is the change, which would be metric[n] - metric[n-1].  I could then divide that by the interval between data points to get the items / second.

Is there a way to do this with a super metric or in vCOps at all?  I'd prefer to do the computation on the metrics in vCOps rather than in the query itself to keep the query a bit simpler.

Thanks,

Jason

[mark_j]

The problem is with the data you're feeding vC Ops. Whereever you're getting it, those metrics shouldn't all be cumulative. That's not the right way to collect performance data in the long run - a short burst test "OK" - long term "no go". Get a solution that feeds vC Ops those metrics in the right format (which most monitoring solutions do).

In terms of converting cum metrics to per-interval, it's not going to happen with a super metric. You can convert a single data point up and down using formulas/etc, however you can't refer to different data points from different collections (e.g. simple moving avg, or referring to prev data point).