Processor yield?
Or
Delay for MilliSeconds per unit of work.
Like opening a barn door.. We added pause logic to halt processes
at key points, and additional logic to perform Panic Stops.
| Previous slide | Next slide | Back to first slide | View graphic version |
Notes:
Multi-processes issues?
I found it difficult to ascertain or track Processor context switches. Various scenarios didn't make any difference to throughput, which was ultimately driven by the IMS transaction response time. I spent a good week fiddling with the code, examining chats and chaing suggestions. Real thread support would have made a difference I'm sure. I settled on a millisecond delay per unit of work to ensure all processes got a kick at the CPU.
As for the debugging, old Cobol programmers were in awe of the ability to dynamically debug and code the image while it was running. One point for our team.