A Holistic Approach to Solving Petascale Problems
C. Fazzino · Proceedings · 2014
As PetaScale high performance computing systems are deployed in the energy sector the full capability of these systems is often not realized. These systems are normally rated by their compute capacity in terms of Rmax , peak FLOPS. This measure says little about how well a system solves a particular problem. Given Supercomputers high cost and the strategic importance of the business decisions they influence, the consequences of delivering an underutilized or over configured system is detrimental to the investment’s value. For Supercomputers to deliver their full value, a comprehensive approach is recommended that considers the problem domain, human factors, software engineering, and system architecture. This paper will discuss how interplay of these factors can work to maximize the value delivered for the supercomputer investment.