Considerations for CICS task history

I recently had a customer with some questions about how OMEGAMON can gather and display CICS task history. Her goal was to be able to have an ongoing history of problem CICS transactions, with the ability to do an analysis after the fact of problem transactions.

When looking at OMEGAMON CICS you have several options for collecting history. First you have the Task History option. This will collect detailed history for CICS transactions to an ongoing task history file that may be viewed in 3270 interface or in the Portal. You can easily filter the display for specific transactions and look for outliers. The main limitation of Task History is that you can only keep what is in the VSAM file used for data collection. It is, by design, a wraparound file. The amount of history kept is a function of the size of the file.

A second option is that you can report from the SMF 110 records created by CICS. OMEGAMON provides the ability to add additional data (like MQ counts, etc.) to these records. The advantage of this approach is that you can keep data for longer periods of time. The down side is that the data must be accessed and analyzed via a batch mechanism, you don't have an "online" method in OMEGAMON to go back in time and look for outliers when using the SMF data.

The third option is to use the Tivoli Portal and to use the Tivoli Data Warehouse (TDW) feature. TDW can snapshot data in the Portal on an interval basis. The interval is definable by the user when you specify the history collection. The data may be kept at the agent (TEMA) level, or optionally sent to the TDW. The advantage of this approach is that it is easy to set up, and the data is readily accessible online, even data going far back. The problem is that the data is purely snapshot in nature. Even on a 30 second interval (as seen in the example), you can have a lot of transactions flow through between snapshot intervals that may not get recorded. It may be that if your goal is to capture outliers that run longer than the snapshot interval, this will work. Otherwise, maybe not.

So to get back to the original question. What if you want history that can be kept on a long term basis to look for outliers and problem trans? That is where maybe you consider other solutions, such as CICS Performance Analyzer or Tivoli Decision Support. What each of these solutions provide is a means to collect the history data and keep it for longer term analysis.