Open for Voting

FEATURE REQUEST - Anomaly Detection Alerting Enhancements

Anomaly Detection using a Machine Learning algorithm was introduced in DPA 12.1.  It works great; if you haven’t explored it, take a look as you may find it’s very useful.  Of course everything can be improved.  Below are two expansions to current anomaly detection functionality that would make it even more useful, particularly for alerting.

  • We use anomaly detection for alerting. However we receive alerts for anomalies on relatively inactive databases because the wait time in a one-hour interval reaches one second, and the norm is 600 or 800 ms.  These alerts have a status of HIGH or MEDIUM so I can't ignore them.  I'm looking for a way to not generate an alert in these situations unless the total wait time exceeds a threshold I set at the database level.  By setting to 5 seconds, 10 seconds, or 30 seconds (whatever is appropriate), I can prevent alerts for these "unimportant" anomalies.  I do need to monitor and alert on these relatively inactive databases which very rarely have an issue because there was an issue last month while I was out on vacation.  And, with this capability, I would eliminate over 90% of our anomaly detection alerts.  Refer to support ticket #00848497.

  • When creating an alert on database wait time anomaly, the minimum execution interval is 60 minutes. Anomaly detection processing executes and updates wait meters every ten minutes based on the most recent 60 minutes.  It would be great if the alert minimum execution interval could be lowered to 10 minutes.  I’d much rather receive the alert after 10 minutes versus 60 minutes to be closer to proactive.  Refer to support ticket #00848505.

Thank you