Self-optimizing resource management in High Throughput Computing Systems

Abstract

High Throughput Computing (HTC) Systems like Condor face the challenge of delivering high throughput in face of rapidly evolving environments. The key to this lies in optimum utilization of available resources and various algorithms have been proposed to achieve this. In mathematical optimization theory, the Simplex algorithm is used to provide optimal solutions to linear programming problems while considering the system's defining constraints. Simplex algorithm has been successfully used over the years to deliver self-optimization to systems requiring maximum output. This work proposes that integrating Simplex algorithm into Condor's resource management framework can instill self-optimizing capabilities in the HTC. The results not only improve the performance with a higher throughput, but also lead to the greater user satisfaction.