关键词: 云计算, 负载均衡, 分层算法, 容错机制, 资源利用率

Abstract: When hybrid dynamic load balancing algorithm is applied to cloud computing, some problems will occur, such as that frequent exchange of sites information leads to low processing efficiency and algorithm is lack of fault-tolerant mechanism. Hence this paper proposed a load balancing algorithm based on multi-layer and fault- tolerant mechanism. The algorithm mixed the advantages of centralized and distributed methods. By organizing neighbor sites, the sites information exchange was controlled within a range of neighbor sites. When site scheduled tasks, it appended the load information of itself and its neighbors to the job transfer request. The method resolved network business and low efficiency of servers which was caused by broadcasting load information frequently. The algorithm achieved load balancing in cloud computing and minimum response time. The method introduced fault-tolerant mechanism to enhance the scalability of cloud system. The experimental results show that the load balancing strategy of cloud computing based on multi-layer and fault-tolerant mechanism is superior to traditional algorithms more than 20% in task distribution time and response time. Besides, it surpasses the traditional one in the stability of the algorithm.

Key words: cloud computing, load balancing, multi-layer algorithm, fault-tolerant mechanism, resource utilization rate