Cost-effective fault-tolerant scheduling algorithm for real-time tasks in cloud systems

Abstract

Scheduling tasks in distributed systems is a classic optimization problem. Cloud computing brings new challenges for traditional scheduling methods due to its characteristics of elasticity. Although the task scheduling problem has been widely studied, there are few heuristics suitable for the cloud environment. In this paper, we propose a cost-effective fault-tolerant scheduling algorithm (CEFT) for real-time tasks in cloud systems. Particle swarm optimization (PSO) is tailored to address the task assignment issue. Primary/backup (P/B) approach is applied to provide fault tolerance for tasks in case of permanent or transient hardware failure. In addition, rescheduling mechanism is put forward to meet the deadline constraints of real-time tasks. Simulation experiments are conducted to evaluate the effectiveness of our algorithm. The results show that CEFT makes good balance between low cost and high deadline guarantee ratio.