Evolving Requirements and Trends of HPC

Abstract

High-performance computing (HPC)high-performancecomputing (HPC) denotes the design, build or use of computing systems substantially larger than typical desktop or laptop computers, in order to solve problems that are unsolvable on these traditional machines. Today's largest high-performance computers, a.k.a. supercomputerssupercomputer, are all organized around several thousands of compute nodescompute node, which are collectively leveraged to tackle heavy computational problems. This orchestrated operation is only possible if compute nodes are able to communicate among themselves with low latency and high bandwidth. In 2004 the ASCI Purple supercomputer was the first to implement optical technologies in the interconnects that support these internode communications. However, research on optical interconnects for HPC applications dates back to the early 1990s. Historically, HPC has been a large driver for the development of short-distance optical links, such as the ones found in today's datacenters (as described elsewhere in this volume). As the number of research areas and industries that exploit HPC is growing, the need for improved HPC interconnection networks is expected to persist. In this chapter we review the requirements of current HPC systems for optical communication networks and we forecast future requirements on the basis of discernible HPC trends.