Understanding resource selection requirements for computationally intensive tasks on heterogeneous computing infrastructure

Abstract

Scientists and researchers face challenges in efficiently configuring their scientific computing tasks so that they can be run in a timely and cost-effective manner. While the increasing availability of different types of computing platforms provides many opportunities to users, it can further complicate the job configuration process. In this paper we present work-in-progress to develop an approach to assist with identifying the most suitable computing platform and configuration for a computational task based on a user’s financial and temporal constraints, using a decision support system. We use Nekkloud, a web-based tool for running computations via the Nektar++ spectral/hp element framework, as an exemplar and build a table that scores a range of properties for four example computing platforms to help select the most suitable platform for a job. We demonstrate our approach using three sample task scenarios.