摘要： 针对云数据中心网络内部流量快速增长导致链路拥塞及全网通信代价过高的问题,提出一种改进的网络感知虚拟机放置算法。根据虚拟机的硬件资源需求,通过种群初始化提高算法收敛速度,采用离散差分变异和交叉操作保证种群多样性。综合考虑通信代价、最大链路利用率、硬件资源约束违反度和链路容量约束违反度4项指标,提出基于ε松弛的多子群精英选择策略,选择最优虚拟机放置方案,增强算法全局搜索能力。仿真结果表明,该算法能够有效降低全网通信代价,并实现网络负载均衡。

关键词: 数据中心, 网络感知, 多目标, 虚拟机放置, 差分进化

Abstract: In view of the problems of link congestion and high network communication cost that result from the rapid growth of traffic within the cloud Data Center(DC),this paper proposes an improved network-aware virtual machine placement algorithm.According to the hardware resource requirements of virtual machines,the population initialization is used to improve the convergence speed of the proposed algorithm.Then,discrete differential mutation and crossover operations are used to ensure the diversity of the population.With a comprehensive consideration of 4 evaluating indicators including communication cost,maximum link utilization,hardware resource constraint violation degree and link capacity constraint violation degree,this paper proposes a multiple subgroup elitist selection strategy based on ε relaxation to select the optimal virtual machine placement scheme and enhance the global exploitation ability.Simulation results show that the proposed algorithm can effectively reduce the network communication cost and achieve the load balancing of network.

Key words: Data Center(DC), network-aware, multi-objective, virtual machine placement, Differential Evolution(DE)

中图分类号: 

• TP393