Roded, You should expect your CPU busy to be higher, but for less time.
For example if the current program runs for four hour using 1 hour of CPU time it will contribute 25% of a single CP to CPU busy. If you cut it into four parallel jobs that run for one hour and use one hour of CPU time the combined jobs will contribute 100% of a single CP to CPU busy. For a batch process reducing elapsed time is the target you are measuring, but this can vary according to how busy other resources are at the time. Your critical path jobs should be set up with a high priority and importance in WLM to minimize variability - the critical path has zero float. When you clone your job to run concurrently make sure that other contention does not saturate other resources. Common files may also have to be distributed among more disk drives and channels, cloned with FlashCopy, or other techniques like Dataset Striping or DLF used to ensure that the full level of parallelism is achieved. Ron > > You can't use elapsed time. Even running the same job over and over > again you will get different elapsed time. > > In fact it is possible that nothing will change. If both jobs are > competing for the same resources (data?) then they could hold each > other > up enough to make no change in elapsed time. > > It would probably better to attempt to tune the application or the > environment. > > It all depends on what the job is doing. > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
