Measure workflow performance

Statistics collection provides performance data you can use to analyze and improve the performance of your system. You can generate statistics for rosters and queues for a Process Engine server in a non-farmed Process Engine system. (In a farmed system, you must use Process Analyzer to generate and view statistical information.)

All work item and queue statistics fall into one of three categories: event summation statistics, sampling statistics, and current statistics.

Category Description Notes

Current statistics

Reflect the state of the running system at the time you request the information.

Current statistics are not stored in the statistics database.

Event summation statistics

Based on the accumulation of events during the specified collection period.

Periods of inactivity can have a misleading effect on event summation statistics. Periods of inactivity can occur when the Process Engine software is not operational due to a system failure or intentional shutdown (for example, for backing up the servers), or when the Process Engine software is operational, but no activity is taking place.

Sampling statistics

Result from an averaging of samples taken at regular intervals during the collection period.

The accuracy of sampling statistics increases as the number of samples increases.

For information on the statistics collected for rosters and queues, see Roster statistics or Queue statistics.

Enabling and configuring statistics

You can enable and disable statistics collection for each isolated region using Process Configuration Console. See Managing statistical information in Help for Process Configuration Console.

You can also specify a statistics collection interval using Process Configuration Console. The statistics snapshot interval defines how often statistical samples or "snapshots" are taken of the work in progress. Note that to make your changes to the statistics snapshot interval take effect, you must restart the Process Engine server. For this reason, we recommend performing this procedure when the software is idle.

Use Process Configuration Console to define the statistics snapshot interval. For step-by-step instructions, see Specify a statistics collection interval.

Methods for retrieving and viewing statistics

You can retrieve statistical data about rosters and queues in a variety of ways.

Process Administrator

Process Administrator provides an easy way to display roster and queue statistics. See View workflow statistics for more information.

vwtool

Use the vwtool statistics command to view roster and queue statistics for a specified time interval.

Java™ and C APIs

You can also programmatically retrieve statistical information about rosters and queues. See Help for Process Development for further information.

How Process Engine stores statistics

Process Engine stores roster and queue statistics in separate tables in the database. A Process Engine server stores statistical information for the rosters and queues in each isolated region running on that server. Each isolated region has its own statistics tables. For example, if a server has three isolated regions, the database contains three tables for roster statistics and three tables for queue statistics.

Statistics database maintenance

The statistics database tables can grow quickly with statistics collection enabled for the isolated region. The system administrator must periodically perform the following maintenance tasks on the statistics tables in the database:

Use vwlog to coalesce statistics and purge statistics data on the current server. See the online help for the vwlog utility for more information.