Availability based Scoring Model for Resource Grouping in Desktop Grid Computing

Abstract

Many scientific projects such as SETI@home are leveraging on enormous numbers of distributed internet-based idle desktop computers to execute their jobs in a Desktop Grid (DG) computing environment. These idle desktop computers (volunteers) are volatile, i.e., they are free to join and leave the DG systems. Besides, the volunteers might also have different resources which are heterogeneous in characteristics and capabilities. Due to the above facts, volunteers in a DG system have different levels of reliability and availability which makes scheduling of jobs more challenging. Thus, this paper presents a grouping method which assigns these heterogeneous volatile volunteers into different levels of reliability groups, according to their availability score. In order to derive the availability score model for the grouping purposes, this study uses statistical analysis of real trace data which are taken from SETI@home project. By having the proper grouping, better assignment of jobs can be achieved in a DG system. Simulation runs of the approach shows that it is feasible to volunteers are able to be grouped using the scoring model.