Background Aggregations

[arunoda]

Currently all metrics received are stored. When processing all of them wiill be counted. This is great for five_secs time resolution.

But if we drawing hourly or daily aggregation, this whole set of metrics slow down the aggregation.

Solution

We can use background aggregation for this. For example.

If we are looking for hourly metrics. We can run a background operation which runs per 1/2 hours and aggregate all the metrics for that our and persist it.

Then we aggregating hourly metrics it simply for this newly stored metrics. It speed up the aggregation in a noticeable manner.

For that we have aggregate for all 4 resolution(sum, avg, max, min) in the background