Dynamic multilevel dual queue scheduling algorithms for grid computing

Abstract

Grid computing is the enabling technology for high performance computing in scientific and large scale applications. Grid computing introduces a number of fascinating issues to resource management. Grid scheduling is a vital component of a Grid infrastructure. Reliability, efficiency (in terms of time consumption) and effectiveness in resource utilization are the desired quality attributes of Grid scheduling systems. Many algorithms have been developed for Grid scheduling. In our previous work, we proposed two scheduling algorithms (the Multilevel Hybrid Scheduling Algorithm and the Multilevel Dual Queue Scheduling Algorithm) for optimum utilization of CPUs in a Grid computing environment. In this paper, we propose two more flavours of Multilevel Dual Queue scheduling algorithms, i.e. the Dynamic Multilevel Dual Queue Scheduling Algorithm using Median and the Dynamic Multilevel Dual Queue Scheduling Algorithm using Square root. We evaluate our proposed Grid scheduling, in comparison to other well known scheduling algorithms, on an SGI super computer using parts of the 'AuverGrid' workload trace. The main purpose of scheduling algorithms is to execute jobs optimally, i.e. with minimum average waiting, turnaround and response times. An extensive performance comparison is presented using real workload traces to evaluate the efficiency of the scheduling algorithms. To facilitate the research, a software tool has been developed which produces a comprehensive simulation of a number of Grid scheduling algorithms. The tool's output is in the form of scheduling performance metrics. The experimental results, based on performance metrics, demonstrate that our proposed scheduling algorithms yield improvements in terms of performance and efficiency. Our proposed scheduling algorithms also support true scalability, that is, they maintain an efficient approach when increasing the number of CPUs or nodes. This paper also includes a statistical analysis of the 'AuverGrid' real workload traces to show the nature and behavior of jobs. © 2011 Springer-Verlag.