Optimizing Cloud Pak System Software Monitoring Server resources

Here are some tips for making the most of your monitoring resources:

  • Cloud Pak System Software Monitoring Portal must run every query you use for different chart and table views. If you use the same query for these views, data retrieval speeds up and you conserve Tivoli Enterprise Portal Server and monitoring server resources.

  • Cloud Pak System Software Monitoring Portal uses queries to specify the attributes to retrieve for a table or chart. After that data is retrieved, you can use the Properties editor Filters tab to filter out any unwanted data from the view. You can use the same filtering capability in the query itself to filter the data before it is retrieved. This pre-filtering means the portal server has less data to retrieve and therefore speeds up the process.

  • You can control the time span shown in a chart or table by adding a filter (through the Properties window) or by using history configurationtime span in the chart or table view toolbar. Using the Filters tab causes the filtering to be done afterthe data has been received at the client. history configurationtime span pre-filters at the server, so only the rows that pass the test are returned, and you conserve network resources.

  • When you create a situation, the interval is set to 15 minutes by default. Set the situation to a longer interval between data samplings to reduce traffic on the monitoring server. This is done in the Sampling interval area of the Formula tab in the Situation editor.

  • A workspace gathers new data samplings when you first open it. You can set the workspace to refresh automatically anywhere from every 30 seconds to once an hour. If you have workspaces set for automatic refresh, set the interval as high as possible to reduce network traffic (requests from the portal client go through the portal server to the monitoring server, then to the monitoring agent, and the information is sent back through the same path).

  • Event data is stored in the KFW tables on the portal server. To ensure the performance of your portal server, you must periodically prune this data. You can do this by controlling how long the event data is kept (using the KFW_EVENT_RETENTION parameter), at what time of the day the pruning begins (using the KFW_PRUNE_START parameter), and when the pruning ends (using the KFW_PRUNE_END parameter). Set these parameters in the Cloud Pak System Software Monitoring Portal Server environment file (KFWENV), as described in the Cloud Pak System Software Monitoring Server Administrator's Guide.

  • A poorly written situation can have a surprisingly large impact on CPU usage on a monitoring server. Using wildcard characters (*) in situations is one of the most common mistakes made in writing situations. Using fields longer than required can also cause additional work to evaluate the situation. Making the changes below have, in some instances, reduced CPU usage on a monitoring server from close to full utilization to nearly completely idle. The CPU usage manifests itself in the kdsmain process. See High CPU usage on the monitoring server for more information.