An Improved Reliability-Driven Task Scheduling Algorithm in Cloud System

Abstract

Task scheduling in cloud computing system is a NP-Hard problem. Cloud system is more complex than distributed system because there are large numbers of servers, so the complexity increasesthe system failures. In this paper, we proposed a reliability-driven scheduling algorithm for directed acyclic graph(DAG) tasks. A two-phase method is approached to solve this problem. In the first phase, we determine both reliability requirement and timing constraint which can be satisfied with the minimum total system cost in cloud system, in the second phase we used greedy algorithm to minimize an aggregation of reliability and constraints time. Experiments results show that the proposed algorithm with high reliability can effectively reduce the total cost compared with the previous work.