Summary
Asia-Pacific Network Operations and Management Symposium
2022
Session Number:TS9
Session:
Number:TS9-01
LFOD: A Lightweight Flow Table Optimization Scheme in SDN Based on Flow Length Distribution in the Internet
Heng He, Zhezhe Peng, Xiaohu Zhou, Jia Wang,
pp.-
Publication Date:2022/09/28
Online ISSN:2188-5079
PDF download
Summary:
Software defined networking (SDN) is a novel network architecture, which decouples the control plane from the data plane which presents many advantages, e.g., openness, management simplicity, etc. In order to support high-speed packet forwarding in SDN, most commercial SDN switches use ternary content addressable memory (TCAM) to store flow entries. Due to the limitations of power consumption, manufacturing cost, and chip size, the capacity of TCAM in switches is limited, which may result in flow table overflow, and is hard to satisfy the performance requirement of the large-scale network. Based on the actual distribution of the flow length in the Internet, we proposed a lightweight flow table optimization scheme in SDN (LFOD). In LFOD, an efficient elephant flow detection algorithm is designed, and it can discriminate the packet whether belongs to mice or elephant flows. Then, a hybrid forwarding mechanism is used to forward flows. As for mice flows, the packet is forwarded to the destination by the controller via packet-out message directly. Regarding elephant flows, the packet is forwarded by installing flow entries on switches. We implement LFOD in the Ryu controller and use realistic Internet traffic traces to evaluate LFOD. The experimental results show that LFOD can significantly reduce the number of flow entries in the flow table, and the workload of the controller keeps relatively low. LFDO can address the issue of limitation of flow table space and make the SDN switch to satisfy the performance requirement of the large-scale network.