|
Abstract:
|
The management of performance variability is an important research area in high-performance computing (HPC). The performance variability is affected by complicated interactions of numerous factors, such as CPU frequency, the number of I/O threads, file size, and record size. In this paper, we focus on the I/O variability, which is measured by the I/O throughputs. To study the performance variability, computer scientists often use full factorial designs to collect I/O throughput data under various system configurations, which becomes prohibitive when the number of factors is not small. In this paper, we propose to use a space filling design to collect I/O throughput data. The advantage of using a space filling design is that it provides good coverage of the design space, many levels for each variable, and good projection properties. We then compare the effectiveness of different design strategies in the setting of HPC performance management. We finally provide some guidelines for computer scientists to collect data using the design of experiments techniques.
|
