Service-oriented, Distributed, High-Performance Computing

Abstract

High-performance computing (HPC) has evolved from a discipline solely concerned with efficient execution of code on parallel architectures to be more closely aligned with the field of distributed systems. Modern HPC is as much concerned with access to data and specialised devices in wide-area networks as much as it is with crunching numbers as quickly as possible. The focus of HPC has shifted towards enabling the transparent and most efficient utilisation of a wide range of capabilities made available over networks, in as seamless a way as that the electrical grid delivers electricity. Such a vision requires significant intellectual and architectural investment. In this article we explore one service-oriented approach for enabling Internet-scale, high-performance applications based on work completed as part of the £250 million UK e-Science programme.