Closed rju closed 2 weeks ago
author André van Hoorn -- Wed, 29 Apr 2015 08:59:48 +0200
During "real-time processing", this should be possible by connecting a time trigger (which already exists). The challenge is during offline processing, i.e., when the monitoring data is read in batch, e.g., from a FS monitoring log, and the analysis time has no relation to the calendar time.
author André van Hoorn -- Mon, 11 May 2015 14:42:48 +0200
Feature implementation (incl. Tests) by Teerat merged into master in changeset:0890f2fdd8ec3a8dcce7221886fa7702c93a73e6/kieker-git. Thanks!
During implementation, Teerat noticed that the existing implementation of TimeSeriesPointAggregatorFilter and its tests need some cleanup/refactoring. Issued a new ticket for this: KIEKER-1370 Done
JIRA Issue: KIEKER-1389 TimeSeriesPointAggregatorFilter should deliver aggregated value in realtime Original Reporter: Teerat Pitakrat
The current implementation of TimeSeriesPointAggregatorFilter delivers the aggregated values only when a new monitoring data arrives. In some cases when there is no monitoring data for a long period of time followed by a new monitoring data, the filter will produce a lot of NaN data points at once. Moreover, these data points can also be very old compared to the newest monitoring data.
It would be nice if the filter can deliver NaN data points in realtime without waiting for a new data point to arrive.