Performance processing for batch jobs

• Bypasses the job to allow normal sequential processing

  If the job does not match any selection criteria, Job Optimizer bypasses the job and provides no performance processing. The job is processed as it would be without Job Optimizer, and analysis and splitting are not performed. You can request Job Optimizer to bypass a job. Request bypass processing to restrict specific jobs when a general solution is used. For example, if you were selecting jobs based on a wildcard representation.

• Analyzes the job during normal sequential processing

  When you request analysis of a job, Job Optimizer does not act on the job, but merely observes the job and records its observations. Job Optimizer studies the job structure–the pattern of EXEC and DD statements. Job Optimizer watches the data activity as each job step executes and determines which steps access each data set and the nature of each access (such as read, write, create, or modify). Job Optimizer observes CPU use during step execution. Job Optimizer saves all analysis data (job history) in the History data set. Splitting does not occur if application behavior is stable or history is not available for the job. Application stability is required for Job Optimizer to safely automate split processing.

• Splits the job steps for parallel processing, targets job steps to images, and pipes data between steps to enable data-dependent processing

  When you request job splitting, Job Optimizer performs job performance processing for jobs that have been analyzed sufficiently. Job Optimizer determines which steps can be run in parallel, and which steps are more CPU or I/O intensive. Job Optimizer targets the appropriate job steps for execution on the available images, intelligently balancing the batch workload across the sysplex. Job Optimizer pipes data between split steps and merges the SYSOUT output and the split step job logs into a single job where the job was initiated.

• Reads JTL (Job Transformation Language)

  Job Transformation Language (JTL) commands are Job Optimizer control cards within a job. These commands provide user-directed job performance processing. Job Optimizer performs the action that is specified by JTL.

This section provides the following topics:

• Information used to determine whether a job can be split
• Determining whether a job can be split
• About data constraints
• About step constraints