Research On Timeout Optimization Of SDN Switch Flow Table Based On OpenFlow

The Internet has become one of the important infrastructures in today's society.The traditional network architecture with TCP/IP as the core is difficult to manage and support the rapid forwarding of a large number of low-latency,high-load data streams in the current network.In recent years,OpenFlow-based Software Defined Networking(SDN)technology has received extensive attention.The software-defined network architecture has a centralized control plane and a distributed forwarding plane,which makes it possible for high-performance forwarding and flexible management networks.In the OpenFlow network,the forwarding of data packages completely depends on the flow entries stored in the forwarding device.As the OpenFlow protocol is upgraded,the matching fields in the flow entry are also increasing,which provides a finer granularity of control for "flow".Since the explosive data flow makes the limited flow table space face a severe test,how to store more entries in a limited flow table space is an urgent problem to be solved.Therefore,this paper presents the optimization problem of switch flow table storage.The main contents are as follows:1.For the problem that the flow table entries are not timely updated or even caused packet loss due to limited storage resources of the flow table during network burst,a flow table timeout optimization method based on load prediction is proposed.This algorithm collects historical newly added data to determine the network status,and then uses the improved algorithm to predict the number of entries that may be reached in the next period.The forecasting value dynamically adjusts the idle timeout period of the flow entries to the newly added table,which vacates the flow table resources in advance for newly added items and reducing the impact of network transmission delay on the data flow.The simulation results show that this method can effectively predict new entries and improve the forwarding success rate.2.When the controller sends out the flow entry,the corresponding timeout time is not set for the characteristics of the data flow itself,which causes the mouse flow entries to occupy a large amount of flow table storage space and then affect the storage efficiency,and theelephant flows are frequently removed by the switch due to the timeout period,causing network thrashing.This paper proposes a dynamic adjustment timeout and entry priority method based on flow identification.By obtaining the counter value of the entries in a certain period,the characteristics of the flow are analyzed,adjusting the timeout period of the entries,and reducing the controller frequently issues items due to frequent deletion of entries.The simulation results show that the adaptive timeout adjustment method of the flow table can effectively save the space of the flow entry and improve the efficiency of the flow table.