IBM Tivoli Monitoring > Version 6.3 Fix Pack 2 > Installation Guides > Installation Guide > Performance tuning > Tivoli Enterprise Portal Server

IBM Tivoli Monitoring, Version 6.3 Fix Pack 2

Portal server parameter tuning

This section provides some information about parameters you might consider editing to improve portal server performance.

The parameters are set in the following files according to operating system:

ITM_HOME/CNPS/kfwenv

For example: C:\IBM\ITM\CNPS\kfwenv

ITM_HOME/config/cq.ini

For example: /opt/IBM/ITM/config

On each occasion maintenance or reconfiguration takes place in your environment these files might be recreated and changes lost and must be reapplied.

For parameter changes made in the portal server configuration file to take effect, the portal server must be stopped and restarted.

KFW_CMW_EVENT_SLEEP

In a complex environment, you might have a number of events occurring simultaneously. This variable specifies a time in seconds the portal server waits between processing similar events. Consider setting this variable to a value of less than 10 seconds if you are experiencing slow performance, such as portal client refresh, as a result.

KFW_EVENT_ASSIST

The Event Assistant is an internal component within the portal server that allows the user to:

  • Associate attachments to events, such as logs

  • Assign ownership to events and transfer ownership between users

  • View events specific to the current user for the convenience of working with events which they own

  • View closed events along with any associated information provided by the user

The Event Assistant creates multiple tables within the portal server database, and processing overhead by the Event Assistant increases the portal server CPU, disk and memory usage.

The Event Assistant is enabled by default. To reduce the processing demands for the portal server, the Event Assistant can be disabled by setting KFW_EVENT_ASSIST=N in the portal server configuration file (KFWENV on Windows, cq.ini on Unix or Linux). Disabling the Event Assistant causes the following behavior:

  • Event records are not written to the portal server database.

  • The ability for any user to attach files to an event is disabled. This is equivalent to all users not having the Attach permission.

  • The My Acknowledged Events view are not updated with events that are being acknowledged.

  • The Event Notes view are not populated. However, to view notes and previously existing attachments, a user can still view notes through the Event Notes and Acknowledgement dialogs.

  • The Similar Events by … view is not populated.

KFW_EVENT_RETENTION

The number of days to keep a closed event. For example, to prune an event 2 days after it is close, specify 2. By default, no event pruning occurs. If the Event Assistant is disabled (KFW_EVENT_ASSIST=N), this parameter is ignored.

KFW_PRUNE_START

The time of day to start pruning data, specified in 24-hour notation (hh:mm). For example, to begin pruning data at 11:00 PM, specify 23:00. By default, no event pruning occurs. If the Event Assistant is disabled (KFW_EVENT_ASSIST=N), this parameter is ignored.

KFW_PRUNE_END

The time of day to stop pruning data, specified in 24-hour notation (hh:mm). For example, to stop pruning data at midnight, specify 24:00. By default, no event pruning occurs. If the Event Assistant is disabled (KFW_EVENT_ASSIST=N), this parameter is ignored.

You can control the size of file attachments for events either at the individual client level or at the monitoring environment level, by changing environment variables in the portal server environment file. Consider changing these if you want to restrict the number of large attachments to be held at the portal server:

KFW_ATTACHMENT_MAX

Use this parameter to control the maximum size of all the files attached to an acknowledgment. Consider editing this parameter if the size of event attachments are too large and are causing network issues, or alternatively you have excess capacity that may be used and attachments have been discarded. Enter the maximum size in bytes, such as 1000000 for 1 MB. The default value is 10000000 (10 MB). If the Event Assistant is disabled, this parameter is ignored.

KFW_ATTACHMENT_SEGMENT_MAX

This parameter enables you to set a size limitation for individual files attached to an acknowledgment. Enter the maximum size in bytes, such as 250000 for 250 KB. The default value is 1000000 (1 MB). If the Event Assistant is disabled, this parameter is ignored.


Topology update processing

The portal server maintains an internal representation of the topology of the monitoring environment, which is called the topology tree. When changes to the topology occur, the portal server updates the topology tree. Examples of topology changes that cause updates to the topology tree include:

The process of updating the topology tree can be CPU-intensive for large environments, and the portal server can cause high CPU utilization for a brief period, depending on the size of the environment and the speed of the processor running the portal server. Performance is also dependent on the number or processors or processor cores. The rebuilding of the physical navigator tree utilizes all processors on the box to improve response time. To minimize these processing requirements, the portal server batches topology updates to be processed as a group. The following portal server environment variables control how topology updates are batched together and affect the frequency of update processing for the topology tree.

KFW_CMW_UPDATE_TOPOLOGY_CUTOFF=xx

This variable sets how long the portal server waits after receiving a topology update for new updates to arrive before updating the topology tree. If a new update arrives, the portal server waits for another cutoff interval for more updates to arrive. If no new updates appear, the update processing is started for topology tree. The default cutoff value is 20 seconds. Changing the value to 60 seconds increases the possibility that topology updates are batched together and can reduce the frequency of topology tree updates.

KFW_CMW_UPDATE_TOPOLOGY_MAX_WAIT=xxx

This variable sets the maximum time the portal server waits when topology updates are arriving before updating the topology tree. If multiple topology updates arrive and are batched together across several cutoff intervals, this variable sets the maximum time the first topology update waits before starting update processing for the topology tree. The default value is 120 seconds. Changing the value to 300 seconds allows for more batching of topology update requests.

Settings made at the client level take precedence over the settings at the monitoring environment level defined here. If you have specified a setting at the client level, then any event variables defined within the portal server configuration are ignored. If you have not specified a setting at the client level, the portal server variables are used. If you have not specified any settings, the default values are used. Figure 1 shows an example of how the topology update buffering algorithm would work using the following values:

Figure 1. Topology update buffering algorithm example

The red arrows above the time line show when topology updates arrive. The blue arrows below the time line indicate when the tree rebuild would start. For the first two topology updates, there is a period of 120 seconds following their arrival when there were no additional arrivals. For the first two updates, the tree rebuild would occur after the CUTOFF interval. For the third topology update arrival, there are multiple arrivals after it, and no period of 120 seconds when no updates arrived. For these updates, the tree rebuild would start after the MAX_WAIT period of 300 seconds expired.

For machines with multiple network adapters, the recommended practice is to specify the IP address for the agent to use in the agent configuration file:

If you cannot update the agent definitions in the time required, as a short-term measure there is a Tivoli Enterprise Portal Server environment variable that you can set to prevent topology updates due to IP address changes. Use of this variable is not generally advised since it prevents the updated information from appearing in the Tivoli Enterprise Portal console.

If agents are not correctly configured and generate excessive Navigator tree rebuilding, you can set this variable to have any discovery of changes or additions of IP address ignored.


Parent topic:

Tivoli Enterprise Portal Server

+

Search Tips   |   Advanced Search