Rescheduling high performance computing jobs based on personalized sanity checks and job problem resolution classification

摘要

A method includes: scheduling a job submitted by a user to be executed on at least a portion of a computing infrastructure; determining a level of sanity check to be performed on the job; performing the sanity check on the job while waiting for the job to be executed; when the sanity check shows that the job will not successfully execute, estimating at least one time required to repair the job such that the job will successfully execute; finding whether the estimated time to repair the job exceeds an estimated time remaining before the job is scheduled to be executed; and when the estimated time to repair the job is found to exceed the estimated time remaining before the job is scheduled to be executed, rescheduling the job such that the estimated time remaining before the job is scheduled to be executed exceeds the estimated time to repair the job.