Monitoring, Version 6.2
The Warehouse Proxy agent is driven by the amount of historical data being collected. The Warehouse load projection spreadsheet can help you estimate both the number of database inserts and the volume of data being collected in the warehouse database.
You can find the Warehouse load projection spreadsheet in the Tivoli® Open Process Automation Library (OPAL) by searching for "warehouse load projections" or the navigation code "1TW10TM1Y" at http://www.ibm.com/software/tivoli/opal.
Figure 4 shows a sample spreadsheet summary created using the tool downloaded from the OPAL site.
The amount of network traffic closely matches the amount of data collected in the Warehouse. The key planning numbers, highlighted in red, are based on sample data. Each Tivoli Monitoring environment is different, so fill out the spreadsheet based on your warehousing needs.
For some high-volume metrics, consider collecting only short-term historical data. For example, if you want to collect process data, there is one row of data per monitored process for every collection interval, which generates a significant amount of data. By retaining only 24 hours of short-term historical data, you do not overload your Warehouse server or network, but you can still perform trending analysis.
If historical data collection is started but a warehousing interval is not set, care must be taken to ensure that the local historical files do not grow indefinitely. This is only for distributed systems. For z/OS®, Tivoli Management Services provides automatic maintenance for the data sets in the persistent data store (the dataset in which the short-term historical data is stored).
The key information in the spreadsheet includes the following data:
In this example, there are 185,600 inserts per hour. Most reasonable database servers can handle this rate without any problem.
In the example, 137 MB of data per hour are inserted. This amount roughly translates to 137 MB of network traffic per hour coming into the Warehouse Proxy agent from the connected agents, and roughly 137 MB per hour of network traffic from the Warehouse Proxy agent to the warehouse database server.
In this example, you expect the Tivoli Data Warehouse database to grow to 143 GB of data. Additional space must be allocated for the database logs. The database does not grow beyond 143 GB because you are summarizing the raw (detailed) metrics and pruning both the detailed and summarized data. The planning spreadsheet helps you determine your warehouse size based on your retention needs.
This warehouse database must be carefully planned and tuned. Where possible separate the largest tables into separate table spaces or data files. By filling in the Warehouse load projection spreadsheet, most users will find that 3 to 5 tables make up the majority of their warehouse data. These tables should be isolated from each other as much as possible so that they have optimal performance.
Multiple Warehouse Proxy agents are required when the number of agents collecting historical data increases above approximately 1500 (to ensure that limits for the number of concurrent connections are not reached). If you use multiple Warehouse Proxy agents, consider running them on the same servers running the remote monitoring servers, and configuring them to support agents connected to this monitoring server. This approach consolidates the Tivoli Monitoring infrastructure components, and limits the number of agents that connect to each Warehouse Proxy agent.
For server sizing, see Locating and sizing the Summarization and Pruning agent.
[ Top of Page | Previous Page | Next Page | Contents | Index ]