CA Gen Upgrade Problem

Subsequent to upgrading to a later version of CA Gen there may be a sudden decrease in the performance of the Access Gen (TSO) batch extract process. What advice can be given to improve the performance of the batch extract?

Solution :

To understand how to take steps to improving the performance of the batch extract, an explanation of how the utility works with the Access Gen (TSO) cache is first required. There are two distinct parts of the batch extract process.

Firstly, there is a Model Structure Cache containing the structure of models stored in the TLIB library. In order to browse a model in Access Gen (TSO), it's structure must be stored in the Model Structure Cache. Initially, each model to be browsed needs to be extracted in it's entirety. Subsequently an incremental process takes over for the changes since the last extract took place. Unchanged models are skipped completely. In general this part of the process is quick to execute after the first extract has taken place.

The second process extracts action diagrams into the ADLIB library. This library is used to save extracting action diagrams from the encyclopedia on demand during the day, which can take a few minutes depending on the size of the diagram and the machine resources available. Note that whether an action diagram is stored in the ADLIB or not, the developer is guaranteed to see an up to date version of an action diagram since if the cached version of an action diagram is out of date or no cached version exists, then the action diagram is re-extracted from the encyclopedia.

From any empty ADLIB, the utility extracts a number of action diagrams up to the limit specified by the batch extract ADLIM parameter. Once the library is up to date, the utility only extracts those which have changed since the last action diagram extract. At this point, the runtime is relatively short. The ADLIM parameter is simply there to prevent excessively long run times when starting with an empty TLIB. The default setting for this is 100.Subsequent to an CA Gen / Advantage Gen upgrade if you started with an empty TLIB, then this would explain why the utility has a great deal of processing to do in order to 'catch up'.

Another point to note is that for the benefit of your developers it is important to note that models are processed in the order in which they are listed in the SYSIN data. Therefore, it makes sense to have the most frequently referenced models (usually the development models) at the top of the list.

If you have done an in-place upgrade of your encyclopedia then you can still use the ADLIB data from the earlier CA Gen /Advantage Gen version. There is no need to extract all the action diagrams again. Similarly, if a DB2 unload/reload or DSN1COPY was used to transfer the data from the old encyclopedia to a new one, the object id's internally would be the same, and the library data can be re-used.