Load balancing algorithm for virtual cluster using fuzzy clustering

Most existing fuzzy-based load balancing algorithms take into account the loads of multi-class system resources, but they cannot truly fuse various loads because the rule based method they utilize is essentially in a combined way. However, the dynamically changing node loads within the virtual cluster cause difficulties in reasonably partitioning the comprehensive loads consisting of multi-dimensional factors. By introducing fuzzy clustering into algorithm design, this paper presents a load balancing algorithm for virtual cluster using fuzzy clustering (LB-VC-FC). To realize the quantization of comprehensive loads, each load of system resources is treated as one dimension of the comprehensive loads and feature weight is attached to each dimension. The improved fuzzy clustering algorithm optimizes distance metric, adds weight constraints and penalty terms, so as to build the specific objective function for load data of virtual cluster. During load balancing, the suitable objective node for virtual machine migration is located based on clustering partition. Experiments demonstrate that the proposed algorithm can achieve fuzzification of multi-class load data and balance loads within virtual cluster.