My requirement is to gather and aggregate metrics for discrete events (execution of specific SQL queries). Events occur at random intervals and the data to be aggregated is number of rows and the time taken per query, which is unrelated to the arrival time of the event.

I am evaluating Open Source packages to gather and aggregate performance metrics, and am looking at things like RRDTool, collectd, statsd, Ganglia and Graphite.

However, all these tools seem to be geared to polling continuous values based on a clock timeline. For example, from the description of rrdupdate:

The update function feeds new data values into an RRD. The data is time aligned (interpolated) according to the properties of the RRD to which the data is written.

I will have on the order of 100000 events per day, so an RRD-like data-reduction architecture is called for, but one where the model is discrete events, not continuous data.

Does such a thing exist, or will I have to write my own? Or, have I missed something that would allow me to use one of these existing systems?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.