数据中心网络中基于蚁群算法的动态多路径负载均衡

针对数据中心网络中等价多路径路由算法（equal-cost multi-path routing，ECMP）无法有效调度大象流而导致流量负载不均衡及易造成网络拥塞的问题，提出了一种基于蚁群算法的动态多路径负载均衡（ant colony algorithm based dynamic multipath load balancing，ADMLB）算法。ADMLB算法首先通过控制器获取网络负载信息，同时检测大象流并标记，然后调用改进的蚁群算法，根据大象流所需带宽选择多路径。实验结果表明，与传统的ECMP和现有流调度算法相比，ADMLB算法降低了链路延迟时间，有效提高了链路带宽利用率。

Dynamic multipath load balancing based on ant colony algorithm in DCN

The traditional ECMP algorithm cannot effectively schedule elephant flows that may cause load unbalancing and network congestion in data center networks. This paper proposed an ADMLB algorithm. Firstly, the controller obtained the load information of each switches and links, detected and marked the elephant flows in the end host. Then, it called the improved ant colony algorithm to solve the multipath problems according to the bandwidth requirements of the elephant flows. Finally, it scheduled the elephant flow. The experimental results show that the ADMLB algorithm reduces the link delay and effectively improves the link bandwidth utilization compared with the traditional ECMP and other existing algorithms.