Comments (2)
  • Add a Comment
  • Edit
  • More Actions v
  • Quarantine this Entry

1 MarvSilverman commented Permalink

To assess how many clones (parallel job streams) are best, I employ expected value analysis to rank each choice. I assign values to categories of elapsed time, CPU delay, InRdyQ, #tape allocations, abend types, DB2 contention factors, etc. and assign risk probabilities to each; then multiply each category by a weighting # to rank its importance. Totaling each cloning choice provides an expected value of the risk/reward benefit to be derived.

2 MartinPacker commented Permalink

Thanks for your input on this. Your approach does, of course, require there to be clones. And to be able to check your assessment it'd be nice to make "recloning" (pardon the neologism) to slide up and down the scale an easy thing to do. I remain convinced automation, in particular in the Scheduler, to be key to this.