IBM Tivoli Monitoring, Version 6.3

Number of worker threads

The Summarization and Pruning Agent creates a pool of worker threads during initialization for performing the summarization and pruning tasks. Each worker thread operates independently of all others, concentrating on one attribute group and its associated warehouse tables. After a worker thread finishes an attribute group, it locates the next attribute group scheduled for processing. If there are no more attribute groups to process, the thread ends, leaving the remaining threads to finish their work.

This threading model provides concurrency while minimizing lock activity, since threads are working on different tables. If there are a few attribute group tables that are considerably larger than the other attribute group tables, the run time of the Summarization and Pruning Agent will be gated by the processing times for the largest attribute group tables.

The number of worker threads can be set in the Additional Parameters tab of the configuration window, or by setting the variable KSY_MAX_WORKER_THREADS in the configuration file (KSYENV on Windows, sy.ini on UNIX or Linux).

The default value is 2.
The suggested number of worker threads is 2 or 4 times the number of processors on the host server. If the Summarization and Pruning Agent is running on a separate server from the warehouse database server, set the value based on the number of processors on the warehouse database server.
Configuring more threads than attribute groups will not decrease the processing time, because each thread works on one attribute group at a time.

Feedback