Research On Flow Table Overflow Mitigation Method For OpenFlow-Based SDN Switches

Software-defined networking(SDN)has become a hot technology in the network field due to its advantages such as separation of control and data plane,programmability,and centralized control.In OpenFlow-based SDN networks,switches use flow table entries to match and forward data flows.For data flows that do not match any flow table entries,switches use PacketIn messages to forward the packets to the controller,which then installs flow table entries to achieve data forwarding.In order to match flow table entries quickly,switches use Ternary Content Addressable Memory(TCAM)to store flow table entries.However,due to the expensive cost and high energy consumption of TCAM,existing commercial switches can only accommodate thousands to tens of thousands of flow table entries.When a large number of new flows burst,switch overflow may occur,which will result in frequent triggering of PacketIn messages,reducing the hit rate of switch flow tables and increasing the controller workload.Therefore,how to develop flow table overflow mitigation strategies to make reasonable use of switch flow tables in limited space is an important issue.The work of this article can be summarized as follows:(1)In response to the problem of limited switch flow table capacity insufficient to accommodate all network requests in a network,which can lead to flow table overflow and subsequently affect the performance of SDN networks,this paper proposes an active perception-based adaptive flow table overflow mitigation method.Based on historical information about the number of installed flow table entries,an adaptive triple exponential smoothing algorithm is designed to predict the number of flow table entries to be installed in the next sampling period.This prediction is combined with the current flow table load to perceive the occurrence of flow table overflow events.When overflow is anticipated,a flow table entry scoring mechanism is employed to prioritize the replacement of low future matching probability flow table entries,allowing the switch flow table space to be rapidly updated to accommodate future network traffic.Experimental results demonstrate that the proposed active perception-based adaptive flow table overflow mitigation method effectively reduces the frequency of flow table overflow events,significantly reduces the generation of PacketIn messages,improves flow table hit rate,and substantially reduces the load on the controller.(2)When deploying flow table entries in open-source controllers,assigning fixed timeout values to data flows with different durations can lead to two problems.Firstly,for short flows,excessively long timeout values can cause them to occupy an excessive amount of flow table space,resulting in the potential for flow table overflow.Secondly,for long flows,excessively short timeout values can cause frequent triggering of PacketIn messages by the switches,thereby reducing the performance of the SDN network.To address these issues,this paper proposes a dynamic timeout mechanism based on load awareness and traffic characteristics.This mechanism monitors the real-time load status of flow tables in switches and employs an AIMD(Additive Increase Multiplicative Decrease)heuristic algorithm to dynamically set the idle timeout for different flow table loads.Through validation using real data center network traffic,the proposed dynamic timeout mechanism effectively reduces the generation of PacketIn messages and significantly improves the hit rate of flow tables in switches,thereby enhancing the performance of the SDN network.