-
Notifications
You must be signed in to change notification settings - Fork 2
Performance Tips and Hints
This page is for performance hints and tips that do not require a full page.
This tip will help the performance of the storage phase for the generic server sensor. It will also help performance of some of the topology agents that clean up unknown servers.
The TADDM 7.3 user's guide documents how to enable the ignore template. Make sure that the ignore template is and always is the last template in the custom servers list. Any other template after the ignore template will not run. If you do this on one discovery server, it applies to all discovery servers in the environment.
To explore unknown servers, which might be desirable during custom server template development, use a development environment. Once you have developed the desired custom server template, define an identical template in production.
Note: Enabling the ignore template may also cause a reduction in relationship discovery. LogicalConnection instances are created based on RuntimeProcess instances that might not be created if the ignore template is enabled.
IIS discovery stores each parameter as a separate instance. This cause cause long storage times during IIS discovery and also hurt ITIC performance. If not needed, it is recommended to disable discovery. The 7.3.0.2 documentation states that this is controlled via a collation property but it does not work on 7.3.0.2. Instead, edit the discovery profile and create new sensor configurations for both IISServerSensor and IIsWebServiceSensor with discoverIISParameters=false. Restart is not necessary.
To reduce the CPU Used by IBM Tivoli Application Dependency Discovery Manager (TADDM) on zLinux, consider the following best practices:
- Reduce the CPU available to the Linux instance. The System z platform offers this opportunity.
- The DWCOUNT in the collation.properties file can be reduced. That reduces the CPU usage and the throughput of the discovery.
When discovering network or storage devices over SNMP, it is a best practice to avoid the session sensor launching. The default behavior is to launch a session sensor and once that fails, move on to SNMP sensors. In the case of SNMP discovery, this is simply wasted effort that causes the discovery to take longer. Take the following steps to avoid session sensor for SNMP discovery.
- Define scope sets for your SNMP targets.
- Define a discovery profile for your SNMP discovery.
- Modify the PortSensor by creating a new one named PortSensorNoPort22 and cloning the original, then edit and set useDefaultPortList to false. Set the portList to 443,1521,445,1741,135,80,23,53,5985,902,7778,5986,389,636. Notice that port 22 is not included.
Now when discovery is run against the SNMP scope set, using the SNMP profile, the session sensor will not be run.
To enhance the performance of the TADDM Discovery Thru ITM feature, ITM APAR IZ63983 is required. Applying this fix will reduce the CPU and network overhead in the ITM environment while a TADDM dicovery is being performed. This enhancement will be available as part of ITM 6.2.1 FP1 and 6.2.2 FP1 IF01.
To get the count for each CI type, run the following database query. You can issue this statement directly against the database or by using bin/dbquery.sh
.
select classname_x, count(*) as count from persobj group by classname_x order by count desc
Best Practices
Sensors
- Supported Target Systems
- Generic Server Sensor (lsof)
- Enhanced CPU Discovery
- Discovering Batch Jobs
- Duplicate Reducer
- VMware Instance Extention
- RHEL Cluster Discovery
- Collecting certs with nmap
- MQ Discovery
- Windows Discovery
- Oracle Discovery
- IIS Discovery
- Ping Sensor
- Collect Config Files Recursively
- Shallow Server Discovery
- Host Storage Sensor