Can cloud computing reach the TOP500?

Abstract

Computing as a utility has reached the mainstream. Scientists can now rent time on large commercial clusters through several vendors. The cloud computing model provides flexible support for "pay as you go" systems. In addition to no upfront investment in large clusters or supercomputers, such systems incur no maintenance costs. Furthermore, they can be expanded and reduced on-demand in real-time. Current cloud computing performance falls short of systems specically designed for scientic applications. Scientic computing needs are quite different from those of web applications|composed primarily of database queries-that have been the focus of cloud computing vendors. In this paper we investigate the use of cloud computing for high-performance numerical applications. In particular, we assume unlimited monetary resources to answer the question, "How high can a cloud computing service get in the TOP500 list?" We show results for the Linpack benchmark on different allocations on Amazon EC2.