PPM Factors to monitor and track

We strongly recommend that you manage performance regularly and proactively so that you can identify areas of poor performance before they become issues for your users. At a minimum, to understand usage patterns and characterize concurrent load during peak hours, monitor and track the items listed in the following table.

Items to monitor	Suggested method
Monitor heap memory usage	JVM verbose garbage collection
Workload and response times	Web log analysis
SQL characteristics (buffers, CPU)	Database reports
Background service execution times	Services performance threshold monitoring framework

For example, suppose your PPM system has 8,000 user accounts, but only 600 to 800 users are logged on to the system concurrently at any given time. By monitoring the patterns of your user community, you can tune the system to meet user needs.

For the PPM database, monitor the following:

• Baseline CPU usage

• Logical I/O per second, which can indicate whether runaway queries exist

• Background services. Use the background services monitor (see Background Services Monitor).

For the PPM application tier, monitor the following:

• System metrics, including:

  • CPU

  • System memory demand (for example, paging or swapping rates)

  • Disk I/O rate

• Heap size

• Thread pool

• Connection pools. Each PPM Server has a distinct set of connection pools. The overall connection demand against the database is the sum of active connections across all nodes in the server cluster. If document management is enabled on your system, you can create additional database connections outside of a node's connection pools.