Tivoli Monitoring, Version 6.2

Locating and sizing the Summarization and Pruning agent

The Summarization and Pruning agent provides the ability to customize the length of time for which to save data (pruning) and how often to aggregate detailed data (summarization) in the Tivoli® Data Warehouse database. Configuration of the Summarization and Pruning agent can have a significant impact on the performance of your Warehouse. Install the Summarization and Pruning agent on the warehouse database server. If that is not possible, ensure that low latency and high network bandwidth exists between your Summarization and Pruning agent and the warehouse database.

Table 2 shows guidance for setting up the Summarization and Pruning agent based on the amount of data inserted into the warehouse database per day. The server sizing is based on the warehouse database and Summarization and Pruning agent running on the same server, and the Warehouse Proxy agent running on a separate server. The assumption is that you have new and fast server-class hardware because the warehouse database server is a high-performing server.

For smaller environments, use a minimum of three dedicated disks because of the high disk I/O requirements of the warehouse database server.

Table 2. Warehouse database server considerations
Data inserted per day Number of CPUs Memory Hard disk storage
0 to 100 K inserts per day 1 CPU 2 GB 3 or more dedicated SCSI drives
100 K to 500 K inserts per day 2 CPUs 2 GB 4 or more dedicated SCSI drives
0.5 to 2 Million inserts per day 2 CPUs 4 GB RAID Array with 4 or more disks
2 to 10 Million inserts per day 4 CPUs 4 GB Multiple RAID Arrays with 5 or more disks. Or, SAN storage
10 to 20 Million inserts per day 4 CPUs 8 GB RAID Arrays with 15 to 20 dedicated disk drives or high-performance SAN storage
> 20 Million inserts per day 4 CPUs 8 GB Multiple RAID Arrays with 20 to 25 dedicated disks or high-performance SAN storage

If your Summarization and Pruning agent is installed on the warehouse database server, configure the agent so that it leaves some processing power for others to perform Warehouse queries and for the Warehouse Proxy agent to insert data. Set the number of worker threads to 2 to 4 times the number of CPUs on the system where the Summarization and Pruning agent is running. Start with 2 times the number of CPUs and then increase the number to see if it continues to improve your performance.

To set the number of worker threads, edit the Summarization and Pruning agent configuration file (KSYENV on Windows®, sy.ini on UNIX® or Linux®) and set the KSY_MAX_WORKER_THREADS to the number of desired threads. This parameter can also be set using the configuration dialog panels.

To assess the performance and throughput of the Summarization and Pruning agent for your environment, you can use the following approach:

  1. Start by enabling historical data collection for a small set of attribute groups which do not generate a large number of rows per data collection interval
  2. Examine the Summarization and Pruning agent Java™ log to see how many detailed records were read and pruned in a processing cycle, and the elapsed time for the processing cycle. Dividing the number of records read and pruned by the elapsed time will give you a rough measurement of the Summarization and Pruning agent throughput (records read and pruned per second).
  3. As long as the elapsed time for the Summarization and Pruning agent processing cycle is acceptable, considering enabling historical data collection for additional attribute groups and repeat step 2.

The processing throughput as determined in step 2 is a rough and easily calculated measurement of the Summarization and Pruning agent performance. Database tuning or additional worker threads can improve the throughput. See the Tivoli Data Warehouse for more tuning information.



Feedback

[ Top of Page | Previous Page | Next Page | Contents | Index ]