High Performance Computing: Current Issues and Future Challenges

Abstract

igh Performance Computing (HPC), the capability of people to achieve computational objectives on existing computers, has made fantastic progress over the past decades. The situation today is excellent in many respects—high performance processors are relatively inexpensive, we understand how to build intermediately-sized parallel computers with scalable performance and cost, and more academic and industry groups than ever before have direct access to HPC resources. For example, such HPC systems are widely used to solve complex application problems in biology, chemistry, physics, mathematics, almost every engineering discipline, and even business and finance. However, current HPC systems have a number of deficiencies that limit their application. For example, the programming environment for parallel computers has improved significantly but still represents a daunting challenge for many potential users. The capability to store and manipulate the large-scale data sets generated by many simulations on today's HPC machines is often prohibitively expensive. And finally, software reuse and interoperability is almost nonexistent in many communities, which forces substantial investments in software development and severely limits the potential benefits of HPC resources. Given this current situation, I would like to briefly address, from the perspective of the JACIC community, what I think the current issues and future challenges are in High Performance Computing.