The role of Analytics in Enterprise Monitoring
By Tom Nedbal, Sr. Principal Consultant, Presales for CA Technologies
“Sometimes you don’t know what you don’t know” said the architect. She was referring to setting Application Performance Management (APM) threshold alerts. Her company has been using CA’s APM solution for 10 years, and they knew from experience which metrics were indicators of impending issues. “If metric X goes over Y threshold, bad stuff happens.” The problem was that their IT environment was constantly changing. Things like new versions of application code, new platform versions and even moving from physical servers to a virtual environment. The old key performance indicators (KPI’s) may or may not be set correctly and there are new bottlenecks unknowingly introduced during every maintenance window.
This graph represents the patterns that may exist in APM metrics today.
CA APM v9.5 introduced Application Behavior Analytics (ABA) as a no charge feature. It installs on its own platform and pulls metrics from the APM Enterprise Manager MOM like a workstation.
- Built on advanced statistical modeling engine tuned to detect simple and complex anomalies
- Notifications on APM WebView Homepage – draw user attention to anomalies
- New “Analysis Workbench” tab allows exploration of multi-source, multi-variant anomalies
- Groups of metrics exhibit unusual interactions, even if thresholds are not violated
- Metrics from multiple sources (agents, passive, third party) analyzed together
- Patterns and interactions revealed in bulk or in isolation
In this example screen shot, Heap Utilization shows that the Garbage Collection spiked and ran out of memory.
If your APM team hasn’t installed ABA in your production APM environment, you may be leaving analytical knowledge on the table!
More information can be found in the “APM -- Application Behavior Analytics Guide” in the CA APM Documentation Library.