Once a job is executing, it might end abnormally because of a hardware, programming, or system error. This might happen any time during program execution. Valuable machine time would be lost if an abnormal end occurred during one of the last job steps of a multistep program or in the middle of a long job step, and execution had to start again at the first job step. There are two ways of avoiding this problem: automatic restart and deferred restart.
For JES2 jobs and JES3 jobs, the checkpoint/restart feature of the system allows a job that ends abnormally to restart either at the beginning of a job step or at a checkpoint within the current step. The programmer submitting the job provides for an automatic restart or a deferred restart.