Availability-based Task Monitoring and Adaptation Mechanism in Desktop Grid System

Abstract

Compliance with the time constraint of the QoS requirement is challenging in the desktop grid environment. Measurements of node behavior indicate the availability of node is cyclic. It is helpful to estimate the execution time of a submitted task on the allocated resource. This paper proposes a novel approach based on available time model to monitor the task execution process and provide running-time adaptation mechanism. Based on this model, a new notion, i.e. Emergency Degree, is proposed to quantitatively indicate the risk of whether the node can complete the task in time. The approach has been implemented and validated on a desktop grid platform - XtremWeb. The evaluation shows the model and mechanism are practical and effective to guarantee the time constraint of the task.