GiAPA – the Operator’s Best Friend

The automated performance analysis tool that got a new job.

GiAPA was originally designed to identify performance problems for jobs running under OS/400. This required collection of rather detailed information – resource usage data for all jobs and tasks every 15 seconds plus call stack and file usage data whenever a job used much CPU.

We expected the data collection to be so resource consuming that it only should run during analysis of a specific problem. But a binary compressed format reduced I/Os so much that GiAPA data collection uses less than 0.1 % CPU and therefore normally runs 24/7.

This opened up for very many additional uses of GiAPA. Containing probably more details than any other tool, GiAPA can document resource usage for all jobs and users, and show statistics for most used programs and files – who, when, what, and how much.

The operations departments are undoubtedly the main everyday GiAPA user, because GiAPA has the data needed to determine the reason for delays, peaks or misuse of server resources and to produce charts with trends per application, resources used by department, most CPU consuming users, variations in load on the LPAR over the month, etc., etc.

GiAPA offers reports and charts depicted by month, week, weekday, date, hour, quarter, minute, or per 15 seconds (allowing to pinpoint exactly when what happened); by job, user, or job type. A simple user exit program allows the user to specify the application code, department and/or division, state or country based on the job ID, in order to provide management with charts showing trends e.g. by application and country.

The analysis and generation of the chart data can be scheduled to run in batch. For all these options you can select which of all the data fields are shown – CPU, 12 different types of I/Os, memory usage, and more…

In addition, GiAPA can send warning messages to the systems operator if a job is looping or allocating excessive memory.