An Efficient Flow Detection and Scheduling Method in Data Center Networks
Zhiyu Liu, Mangui Lian, Jin Guo, Guangchao Wang · The 2nd International Conference on Computing and Data Science · 2021
Nowdays Datacenter networks (DCNs) scale is rapidly increasing because of the widely deployment of cloud services. Thus, a large amount of data needs to be frequently interchanged among thousands of servers. In this environment, a key and challenging issue is to keep the traffic loadbalanced. In this letter, we propose a port forwarding load-balanced scheduling (PFLBS) approach for Fat-tree based DCNs. Firstly, we define a port-based source-routing addressing scheme, which simplifies the switch's functionality and makes the routing lookup operation unnecessary. Secondly, we apply the addressing scheme to the end hosts and formulate the PFLBS problem. PFLBS propose an efficient algorithm for multipath selection to obtain optimal paths and designs a trigger mechanism to split large flow implement scheduling data flows dynamically. The experiment results indicate that our PFLBS approach has better performance compared with the methods ECMP and Hedera and decreases the flow completion time and improve the average throughput significantly.