Using Moldability to Improve the Performance of Supercomputer Jobs

In most parallel supercomputers, submitting a job for execution involves specifying (i) how many processors are to be allocated to the job, and (ii) for how long these processors are to be available to the job. Since most jobs are moldable (i.e. there is a choice on how many processors the job uses), the user typically has to decide how many processors to ask for a given job, and estimate the request time accordingly. In this paper, we show that the request that submits a moldable job can be automatically selected in a way that often reduces the job’s turn-around time. More precisely, we introduce and evaluate SA, an application scheduler that chooses, on behalf of the user, which request to use to submit a moldable job. The user provides SA with a set of possible requests that can be used to submit a given moldable job. SA estimates the turn-around time of each request based on the current state of the supercomputer, and then forwards to the supercomputer the request with the smallest expected turnaround time. The conditions under which SA is studied cover variations on the characteristics of the job, the state of the supercomputer, and the information available to SA. The results show that SA often improves the turn-around time of the job under a variety of conditions.