Acceptance of waiting times in high performance computing

Abstract

In high performance computing, users submit computing jobs to a parallel computing infrastructure. Ressources for jobs must be allocated according to the job’s requirements. At times of high workload, waiting times are unavoidable. Hence, we investigate user acceptance of waiting times in high performance computing. We analyze data provided by Questionnaire for User Habits of Compute Clusters (QUHCC) among high performance computing users (n = 24). On the one hand, the results indicate that increasing processing times of jobs lead to greater acceptance of waiting times. On the other hand, the percental difference between processing times and waiting times decreases for increasing job lengths. We suggest that scheduling strategies should respect our findings.