Choosing between segmented and nonsegmented programs

When you are deciding whether to design your program as segmented or nonsegmented, you should be aware of two issues.

The first issue is the effect of the transaction on contention resources, such as storage and processor use. The second issue is the effect on exclusive-use resources, such as records and recoverable data sets, recoverable transient data queues, and enqueue items.

Nonsegmented programs have a high impact on storage because they run longer than the sum of the transactions that are in an equivalent segmented program. However, processor overhead is lower because only one program is started instead of one for every transaction.

A nonsegmented program retains exclusive use of resources for a longer period of time, unlike the equivalent segmented program. For this reason, segmented programs are quicker to respond, but for recovery and integrity considerations, you might prefer a nonsegmented program.

If you have forms in called programs or need to lock the database during a converse, you should design your program to run in nonsegmented mode.

The following list contains considerations for segmented and nonsegmented programs:
  • Segmented mode uses more processor time because CICS® spends more time initiating and ending transactions.
  • Nonsegmented mode uses more virtual storage because transactions are still active during user think time. However, with dynamic transaction routing CICS® can automatically start another region and send transactions to the next region when the first region is constrained.
  • Nonsegmented mode can also use other resources such as locks in the database during user think time. (This can be solved by setting the system variable converseVar.commitOnConverse to 1.)
  • CICS® accounting and security is less granular with nonsegmented transactions because you have a few large transactions, rather than a lot of small ones.
  • CICS® shutdown can be more difficult with a lot of nonsegmented transactions. You might have to end transactions before you can shut down because someone is out for a break in the middle of a nonsegmented transaction. (This can be solved by having transactions time out if the user has not pressed Enter after a specified time).
  • Programming nonsegmented programs can be easier because you can do the following:
    • Use text forms in called programs.
    • Hold locks and cursor position in the database over a converse.
  • Only segmented programs can be generated for the IMS™ environment.

When running a program on a system where storage contention is not a problem, a good compromise between running in segmented or nonsegmented mode is to run in nonsegmented mode with the converseVar.commitOnConverse variable set to 1. This approach forces a commit at every converse from the main program, and it has the good performance characteristics of nonsegmented mode, but it does not hold file or database resources during user think time.