PCS workload management process

This section explains how PCS is designed to work once the setup is complete. It assumes the environment is ready and available, i.e., CA 7, JES2, and BMC ThruPut Manager started tasks are all operational; the AMD has daily production cycles specified; and the battle plan is auto-activated.

Here are the highlights of the process for a job belonging to an application in monitor mode. Where the process is different for a job belonging to an application in managed mode, this is noted.

• CA 7 database is updated for the day.
• The PCS Forecast Collection Utility is run based on the updated CA 7 schedule.
• The battle plan is created. PCS determines the end time for all paths and milestone jobs, and the duration of all the jobs. From that it determines the expected start time (that's where the CA 7 definition of predecessor jobs and the PCS experience data comes in). Lastly, it determines the initial critical path and the initial urgent path for each job tree.
• The battle plan is activated.
• CA 7 places jobs in the REQUEST queue to await submission. (When such a job is not known by the PCS Forecast Collection Utility, it is referred to as an ad hoc job and acts as a single-job tree).
• PCS is notified when the job has entered the REQUEST queue and later again when it is submitted to JES2.
• As is usual with BMC ThruPut Manager, the job is analyzed and processed by the JAL processor when it is submitted. JAL processes the job adding any binds, limits, messages and so on. All SLM statements are ignored for jobs belonging to an application in managed mode.
• A job belonging to an application in monitor mode is not processed in the PS queue by PCS. Rather it reverts to its non-PCS default processing, which means it is processed by one of:
  • JES2 if the submitting job class is not selectable by BMC ThruPut Manager. (The JECL option JECLEXEMPT also influences whether the job is analyzed or not.)
  • JAL-assigned job class if the submitting job class is selectable and JAL processing does not assign the job to a Service Group.
  • BMC AMI Ops Automation for Batch ThruPut, in the General Services queue, if JAL processing assigns the job to a General Service Group;
  • BMC AMI Ops Automation for Batch ThruPut, in the Production Services queue, if JAL processing assigns the job to a Production Service Group.
• A job belonging to an application in manage mode is "processed" by PCS in the PS queue. PCS subcontracts SLM to manage the job according to the PCS parameters, which it may be updated as the workload progresses. SLM may change the Service Class, give preference to a job with respect to resources, or change its position in the queue.
• PCS keeps track of whether a job is on an urgent path, whether it's going to be late, whether it's a culprit for making a downstream job late, and so on. It sends alerts if one of 17 workload events occurs.
• PCS updates its displays for this job, its path, tree and application, as well as any affected jobs. The PCS job display includes an integrated view of JES2, CA 7 and BMC ThruPut Manager status and information. BMC ThruPut Manager's UDF (User Display Facility) provides further runtime information on each job.
• As each job ends, PCS
  • Updates the status fields,
  • Updates "expected" values with "actual" values,
  • Updates the Experience Collection File,
  • Updates the SMF Record Collection file, and
  • Generates the Job Summary Report as part of the job output, if the appropriate option is set.
• The production cycle ends and the battle plan is closed. Normally a new production cycle is starting with its battle plan. Any outstanding jobs are inherited by the new battle plan.
• Later, the datacenter runs a utility to transfer the Experience Collection File information for that production cycle to the Experience File for aggregation with the information from previous cycles, readying the Experience File for use with the next battle plan.

Ideally, during this process Production Control operators simply sit back and let PCS manage the workload in the optimal manner to best complete the workload on time. However changes in circumstances, priorities, and so on do occur. You have full control to make any adjustments as needed. See Runtime View for more details.