Toward Smart Scheduling in Tapis

The Tapis framework and APIs enable a wide range of research and scientific discovery by automating job execution on remote resources, including HPC clusters and servers running in the cloud. While Tapis simplifies interactions with cyberinfrastructure (CI), users still need to specify all details of a job they want to run, including the system, queue, node count, and maximum run time, among other attributes. Moreover, system and application resources must be defined in Tapis before a job is submitted. It can be a challenge for scientists to provide exact configurations and resource requirements for their jobs, which if incorrectly specified can delay their work.In this paper, we investigate developing an intelligent job scheduling capability in Tapis, where job configuration can be partially automated and computational resources can be dynamically provisioned by Tapis. We outline an architecture for such a feature and identify a set of core challenges to be solved. Then, we focus on one specific challenge: predicting queue times for a job on different HPC systems and queues, and we present two sets of results based on machine learning methods. Our first set casts the problem as a regression, which can be used to select the best system from a list of existing options. Our second set frames the problem as a classification, which allows us to compare the use of an existing system with a dynamically provisioned resource.