IBM Tivoli Monitoring > Version 6.3 Fix Pack 2 > Installation Guides > Installation Guide > Performance tuning > Agentless Monitoring

IBM Tivoli Monitoring, Version 6.3 Fix Pack 2


Adjust the thread pool size, refresh interval, and cache time-to-live parameters for your environment

The main thread pool parameters are CDP_DP_THREAD_POOL_SIZE and CDP_DP_REFRESH_INTERVAL.

For a complete list of Agentless Monitoring parameters, see the Agent Builder User's Guide. The following are some of the main parameters:

CDP_DP_THREAD_POOL_SIZE

Controls the number of threads in the thread pool.

CDP_DP_REFRESH_INTERVAL

Controls how often the thread pool collects the data for each attribute group. The refresh interval should also be correlated with CDP_DP_CACHE_TTL. Both are specified in seconds, and default to 60. If the refresh interval was stretched out to, for example, 300 while the cache ttl stayed at 60, then there would be a 4 minute period during which the cache was considered expired, and any queries that came in during that window would cause an on-demand collection to happen. The best practice is to keep these two values the same.

To determine if the value for CDP_DP_THREAD_POOL_SIZE is correct, check the Thread Pool Status attribute group and look at the Average Queue Length and Average Job Wait fields, and the Intervals Skipped field of the Performance Object status. If these numbers are getting large, and the processor load on the system is in the acceptable range, then the size of the thread pool can be increased. One thread per subnode might be a good starting point, although if there are fewer than 60 nodes, there is probably no reason to decrease below the default 60 threads. If the load on the system is too high, or if the Average Queue Length, Average Job Wait , and Intervals Skipped do not decrease, collecting the data less often can decrease the load. The default refresh interval is 60 seconds, so increasing that number, and the CDP_DP_CACHE_TTL as well, can decrease the overall load.


Parent topic:

Agentless Monitoring

+

Search Tips   |   Advanced Search