Productivity Metrics and Models for High Performance Computing

Abstract

Historically, high performance computing has been measured in terms of peak or delivered performance, and to a lesser extent to performance to cost. Such metrics fail to capture the impact on the usefulness and ease of use of such systems. Productivity has been identified as a new parameter for high end computing systems that include both delivered system performance and the programmability of the system. System productivity is directly affected by many factors contributing to the achieved performance, the speed with which users construct application programs, and the availability of the system to perform user applications. This paper explores the concept of productivity as a quantifiable; parameter through a series of analytical models and considers the factors that contribute to it.