Perf Minute: Batch Program Scheduling

What does your enterprise analyze when scheduling batch programs to run in your scheduler (Redwood, UC/4-Automic, IBM Workload, Control-M, et al)? Does your batch team consider all PE factors when devising their run schedule?

Often, Batch teams look at when jobs need to start and just hope they finish on time. But there are many factors that determine the optimal use of computing resources, dictating when those jobs will finish and what impact they will have on other jobs.

Factors that should be considered are:

1. CPU Cycles: How CPU-intensive is the job (if production statistics are not available - i.e. it's a new job - analysis must be conducted)?
2. Volumetrics: What volume of data is the job expected to process (how many lines, IDOCS, records, files, etc per metric [hour, day, week])?
3. Dependencies: What does the entire workflow or job chain look like and where does this job fall in that chain?
4. SLA / Completion Window: Does this job have to complete by a certain time? If so, exactly when and why (from a business lens)?

These are just some of the key factors that must be evaluated when building a job schedule. These factors help determine what jobs should run together or how you can stagger the jobs out to most efficiently utilize computing resources, while meeting business requirements.

In addition, this valuable information, once built, will be paramount in building a good Batch Performance (PE) Test Plan to run these jobs in your staging environment prior to going live.

Is your enterprise doing a good job of factoring in the most important details when devising their batch schedule? Are there other key factors that you find helpful when planning your schedule?