A Novel Fault Tolerant Volunteer Selection Mechanism For Volunteer Computing

Abstract

Volunteer Computing (VC) is an exciting discipline that has allowed the use of millions of volunteers connected through internet to participate in any project by donating their processing cycles and memory. VC systems have lot of potential but also possess few challenges such as volunteers are not managed centrally, they can submit erroneous results (intentionally or unintentionally) or can quit at any time even while executing assigned tasks. The simple solution is to select volunteer carefully keeping in view the past performance and other measures. In this paper, we have proposed a fault tolerant framework for selecting the best available volunteer. The proposed framework consists of a volunteer grouping and selection mechanism based on the collective impact of processing cycles, memory, spot checking and makespan statistics. To achieve fault tolerance we have also proposed a dynamic replication mechanism that generates the number of replicas of a particular task by analyzing volunteer’s behavior.