Research On Scalable Routing And Flow Table Resource Optimization In Software Defined Networks

Software Defined Networking (SDN) is a new networking architecture, which aims to separate control plane function from network forwarding devices. SDN is comprised of a centralized controller and SDN switches. In order to obtain better performance, the controller is responsible for all the control functions, while SDN switches just retain simple data forwarding function. In an OpenFlow network, controller maintains a global view of this network, and it determines the forwarding behaviors of switches via standardized OpenFlow protocol.Scalability is one of key problems in SDN. To alleviate this problem, we proposed a new routing scheme for large-scale OpenFlow networks, and tried to utilize flow table resource effectively.(1) When a new flow arrives, the controller computes global optimal routing path for this flow and adds corresponding flow entries into all switches along the path. Obviously, the amount of messages exchanged between controller and switches grows with the size of network. As a result, the poor scalability of traditional OpenFlow routing scheme becomes more and more obvious.(2) In OpenFlow networks, flow table are mostly built from Ternary Content Addressable Memory (TCAM) to achieve fast looking up and forwarding performance. Additionally, a single controller only has the ability to process a limited number of packet-in events. As a result, flow table and controller processing resources should be used effectively. On the other hand, as Instant Messaging (IM) applications usually generate too many different flows and idle_timeout has effect on the two limited resources, we try to obtain the effective idletimeout value for IM. The main contributions of this article are as follows:(1) We propose SRSO, a scalable routing scheme for large-scale OpenFlow networks.When routing a flow, SRSO only adds flow entries into a part of switches (not all switches) along the path. Comparing with traditional OpenFlow routing scheme, SRSO achieves better performance of scalability through decreasing the amount of exchanged messages destined to the control plane. Furthermore, we develop components on POX controller to implement SRSO and test its feasibility in real environment. The simulation results show that our proposed SRSO can obtain better scalability performance in large-scale OpenFlow networks.(2) Firstly, based on ON/OFF traffic model, we explore the effect of idle_timeout on flow table and controller processing resources respectively, and then define corresponding costs of the both resources. After that, we derive the lower and upper bounds of idle_timeout value considering the constraints of limited flow table and controller processing resources. Furthermore, we formulate an optimization model to calculate the effective idle_timeout for IM. Finally, we note that the simulation results are consistent with our theoretical analysis, which verifies the effectiveness of our model.