Identifying High Resource Consumers
IBM Lab Services helps you optimize your system performance by targeting greedy processes
Illustration Courtesy of Shutterstock.com
Over the years, I’ve had the opportunity to review the system and application performance for many i5/OS* and IBM i customers. All of them are concerned about time, and I’m sure you’re no different. That’s why it’s so important to ensure your performance-optimization efforts are directed at the processes with the most potential for improvement.
The analysis steps I’ll outline here should help identify the processes consuming the most processor resources and performing the most I/O on your system.
Collecting Performance Data
Collection Services is the primary collector of system-management data. It provides continuous data collection with minimal overhead. Most customers collect data at the default interval of 15 minutes. You may want to reduce this interval to five minutes or less to collect more specific information during your performance analysis. The collection-interval setting can be modified through the Configure Performance Collection (CFGPFRCOL) command or the Navigator for i interface. If your system isn’t configured to automatically create the database files, you may need to use the Create Performance Data (CRTPFRDTA) command to generate the database files from the management collection (*MGTCOL) object.
Accessing the Data
The Collection Services data is stored in multimember files often located in library QPFRDATA or QMPGDATA. Each collection is stored as a separate member. For customers using the default collection settings, a collection represents 24 hours of data. The following methods can be used to access a unique member of a file using SQL (replacing “collection” with the name of the collection or database file member).
- Override With Database File (OVRDBF) command:
- SQL ALIAS:
CREATE ALIAS QTEMP.JOBL FOR QMPGDATA.QAPMJOBL ("collection")
To switch to another collection member, first use the Delete Override (DLTOVR) command or DROP ALIAS statement. My personal preference is to use the “Run SQL Scripts” interface of Navigator for i with one DROP ALIAS statement and CREATE ALIAS statements for each collection member being analyzed.
Several files with many valuable performance metrics are available in the Collection Services database. For our purposes, we’ll concentrate on just a few metrics from two of the files, QAPMJOBL (job-level performance data) and QAPMSYSTEM (system-level summary performance data).
For a list and description of all of the available metrics in all files, review the “Performance data files” topic in the Information Center.
I like to use a simple system-level summary of a few key performance metrics as a starting point. The SQL statements in Figure 1 (left) summarize processor consumption as well as the rate of file open and logical I/O (read and write) operations for each interval in the collection. The formatting of the date and time allows for ease of integration into a spreadsheet for reporting.
The next step is to drill down into the processes consuming the most processor resources and performing the most logical I/O during those intervals.
Search our new 2013 Buyer's Guide.
Web Exclusive | My experiences with IBM i SAP and SSD
Web Exclusive | How to better utilize your available power