基于模糊离散粒子群的数据中心负载均衡研究

Research on load balancing in data centers based on fuzzy discrete particle swarm optimization

针对传统数据中心网络流量调度策略可能会引发大象流冲突,从而导致网络链路堵塞和负载不均衡问题,提出了一种模糊逻辑和离散粒子群相结合的数据中心负载均衡机制。该研究充分利用软件定义网络集中控制与全局视图特性,首先基于重定义的离散粒子群优化算法对大象流进行筛选;然后设计了一个基于模糊逻辑的综合评价模型,模型根据候选路径的特定参数生成路径综合评分;最后选择综合评分最高的路径转发大象流。实验结果表明,与其他算法相比,本文算法网络吞吐量提高了3.05%,链路利用率提高了8.82%。

The traditional network traffic scheduling strategies of data centers likely causes potential elephant flow conflicts,which can lead to network link congestion and load imbalance issues.In order to solve the problem,a data center load balancing mechanism combing fuzzy logic and discrete particle swarm optimization(FD_PSO)is presented.This research fully harnesses the centralized control and global view features of software-defined networking and firstly filters the elephant flows based on redefined discrete particle swarm optimization(RPSO).Then,a comprehensive evaluation model based on fuzzy logic is designed.This model generates a composite score to the candidate path based on specific parameters,and selects the path with the highest composite score for forwarding elephant flows.The experimental results indicate that,compared to the other algorithm,the algorithm of this study increased the network throughput by 3.05%and improved the link utilization by 8.82%.