Performance monitoring tasks

The following table lists maintenance tasks that a System Administrator needs to perform in the Performance Monitoring area. For information about other areas, see Areas of maintenance tasks.

Event or frequency

Task

Automatable

Daily Review the server log for performance messages Yes
Daily Analyze dbdicts containing log information Yes
Weekly Check view and favorite creation No
Weekly Check the usergrid table for changes to user’s record lists Yes
Daily Monitor system reports for performance related information Yes
Daily Performance monitoring tasks Yes
Quarterly Monitor the dbdicts purging progress Yes

Review logs for performance messages

This task includes the following subtasks.

Review the server log for performance messages

The Service Manager server log files contain a lot of information about system availability and performance.

Review the output produced by alert messages and proactively activated trace parameters. The alert log messages provide information about potential performance issues, or other specific situations that require attention.

Analyze dbdicts containing log information

Service Manager maintains multiple dbdicts with log information for different purposes. Among them, the reportLogs dbdict is important for performance monitoring.

It is important for the System Administrator to review the performance log file and see if a task is slow. Each reporting query may take a different amount of time to execute, and therefore manual review is required.

Look for report queries that are often more complex than other queries and are expected to take more time for execution. It is therefore not possible to define a one-fits-all threshold.

It is important to identify long-running reports, and learn their business requirements and implementation. Based on this knowledge, the System Administrator can decide on how to act on excessive report execution times or the trends that are identified.

Check view and favorite creation

If users are given the ability to create views and favorites (public or private), the System Administrator will need to put into place a process where the administrator can check the records in the inbox table regularly to ensure that the queries and sort sequences used in the views and favorites are properly indexed. If a sort sequence is specified, the fields used in the sort sequence should be part of the index.

Additionally, if the administrator finds multiple users creating the same private inbox, the administrator needs to consider creating a public inbox for all users in a query group (all users that have this inbox created). By doing this, the administrator will be able to keep the number of records in the inbox table under control. For performance tuning of queries, refer to Performance of dbdicts and tables.

Knowledge required for system maintenance > Performance of dbdicts and tables.

Check the usergrid table for changes to user’s record lists

Check the usergrid table to see what users are saving as their qbe formats. If you find that many users have added the same field to a particular QBE, you may want to consider adding the field to the default QBE format. Remember to delete the appropriate records from the usergrid file once you have modified the format.

Monitor system reports for performance related information

Monitor the following system report on a daily basis.

Parameter Description

sm - reportcache

If the average slot depth exceeds 20, consider reconfiguring the shared memory cache size by the cache_slots parameter. For more information, see Startup parameter: cache_slots.

Monitor the server’s performance statistics

Check for memory usage and CPU usage peaks to ensure smooth operation. Over time, with more concurrent users logging in, it may be necessary to add more memory or more CPU, or adding more servlets to serve the increasing user community.

Monitor the dbdicts purging progress

Data, queue, or log dbdicts containing a large number of records can impact system performance, and therefore archiving and purging should be configured.

Monitor the number of records in these dbdicts frequently to supervise cleanup activities.