A static benchmarking for grid scheduling problems

Abstract

Analysis of algorithms for Grid computing systems before deployment in real Grid infrastructures is an important issue in Grid computing domain. Due to the complexity of real Grid systems, assessing performance analysis of optimization algorithms such as scheduling algorithms, is in general difficult, costly and time consuming. Benchmarking and simulation are two most used alternatives for analyzing optimization algorithms in Grid systems before deployment. In this paper we present a static benchmarking for scheduling problems in Grid systems. The benchmarking has been generated using the HyperSim-G Grid simulator and captures several types of Grid systems based on combinations of different machine and task types. Instances have six different sizes ranging from tiny (32 machines/512 tasks) to extra large size (1024 machines/16384 tasks) and are grouped according to machine and task types. The benchmark suite, consisting of about 720 instances, is offered through a web page