There were significant changes to the IBM i performance tools in the 6.1 release. The ability to collect additional types of performance data, management of this data, and a graphical interface for viewing and analyzing this data were added. I often speak on performance topics, but I have not written much about them. Since there is much to cover, this will be the first in a series of articles. A good place to start is to review the performance data collectors available on i.
Beginning with the 6.1 release, there are four different types of performance data you can collect; two of them newly integrated into the operating system with that release. IBM i has the richest set of performance data compared to any other operating system; I believe it is a substantial differentiator, giving i users greater insight into how their machines are functioning.
Collection Services has been part of the operating system for many releases. Most administrators should be aware of it and using it regularly. Collection Services collects performance data from throughout the operating system on an interval basis and is designed to be running continuously. The overhead for data collection is not significant, but you do need to carefully manage the data that is collected. The major change with Collection Services in the 6.1 release is that it was changed to always be running by default.
Collection Services is used for day-to-day management of your system's performance; you use it for your baseline performance data collection to more quickly determine if changes have occurred that affect performance. It's the underlying data for Management Central System Monitors, it's the data that is sent to IBM if you use PM for Power Systems, and it's also the data that's used for trending and capacity planning purposes.
Performance Explorer, commonly referred to as PEX, has been part of the operating system since the V3R6 release. PEX has three modes of operation and can collect statistical performance data as well as trace data. Today, PEX is generally used for two reasons: 1) as a performance collector to aid with detailed application performance analysis, and 2) to collect detailed performance for diagnostics on the most difficult performance problems. PEX used to be used much more, but the development of the Job Watcher performance data collector has changed that.
Job Watcher first appeared
around the V5R1 time frame. Job Watcher was initially developed by the
Rochester Support Center as a tool to aid performance diagnostics. Job Watcher
and the graphical user interface used to analyze the data proved to be very
powerful for performance diagnostics. The iDoctor offering was the first place
where Job Watcher was made available to our clients. Since Job Watcher was so
effective at performance diagnostics, it was built into the operating system
with the 6.1 release.
Job watcher can collect much of the same information that Collection Services does, but it collects much more; call stacks, SQL statements, wait information (see i Can ...Tell You Why You Are Waiting for more information about waits), activation group information, Java information and more. In addition, while Job Watcher uses a sampling technology similar to that of Collection Services, Job Watcher has the option to sample as fast as possible, which can provide data similar to a trace.
Job Watcher is the most commonly used performance data for detailed diagnostics. The Rochester Support Center performance analysts, as well as the performance consultants in IBM Lab Services generally use Job Watcher. Nearly every performance problem can be resolved with Job Watcher data.
Disk Watcher is a specialized performance data collector that collects data specifically for disk operations. Its use is intended for detailed diagnostics of disk performance problems or for the use in optimizing disk I/O in applications.
All of these performance data collectors have a set of commands that you can use to define the data you want to collect, commands to start and end the data collection, and commands to manage the data. There are graphical user interfaces that I'll cover in the future as well.
A prior blog article on Find All my Performance Collections in One Place reviews the Collections task in the Navigator for i Web console. The Collections task can be used to see all of your performance collections, regardless of what type, in one consolidated interface.
As I noted at the beginning of this article, IBM i's performance data collectors are second to none; the combination of Collection Services for day-to-day performance management and Job Watcher for detailed diagnostics provide a powerful combination of tooling that can’t be beat.