182 IBM Tivoli Monitoring: Implementation and Performance Optimization for Large Scale Environments
5.8 Query optimization
This section discusses tuning the queries that are processed to display the
tables, charts, and graphs that make up workspace views within the Tivoli
5.8.1 Query processing
The query assigned to a chart or table view requests data from a particular
attribute group. It executes when you open or refresh the workspace. Queries
make up the processing load for on-demand data collection. You can reduce the
frequency and amount of data sampling by:
Customizing the query to filter out unwanted data. This reduces the number of
selection criteria (rows) and attributes (columns) collected.
Applying the same query to other views in the workspace. This reduces the
number of data samples required: one query uses a single sample for multiple
Disabling automatic refresh of workspace views or adjust the refresh rate to
longer intervals. This causes Tivoli Enterprise Monitoring Agent data to be
collected less frequently.
Consider how you wish to display the data returned by a query. A graphical
view workspace may require significantly less data compared to a table view,
because it uses only the truly numeric data and leaves out the character data.
The following general recommendations and observations should be considered
Some attributes are more expensive to retrieve than others. One expensive
column in a table will make any workspace view or situation that references
that table more expensive. An example of an expensive attribute is one that
must run long storage chains to determine its value, such as using a process
table to look for looping tasks. Where possible, ensure you only retrieve
attributes that are required for the monitoring process.
Column function (such as MIN, MAX, AVG, and so on) requires
post-processing of the query results set after data is returned to Tivoli
Enterprise Monitoring Server.
Note: Do not confuse custom queries with view filters from the Filters tab
of the Query Properties editor. View filters fine-tune the data after it has
been retrieved by the query and do not reduce network traffic, data
collection processing, or memory demands.