The Research On Elastic Energy-saving Storage And Accelerated Lookup Architecture Of OpenFlow-based Flow Table

SDN technology based on OpenFlow protocol has become increasingly mature after ten years of development and evolution,and has been widely employed in large-scale network scenarios such as WANs and data centers in recent years.OpenFlow switches usually adopt Ternary Content Addressable Memory(TCAM)flow tables to forward packets quickly in SDN data plane.However,capacity of TCAM is limited and its lookup process consumes a lot of energy,and large-scale deployment of SDN network will generate a large number of concurrent flows.Frequent lookups of TCAM will bring a large number of energy consumption,which is not conducive to development of SDN green network.Meanwhile,scale of flow tables is becoming larger and larger with continuous evolution and development of OpenFlow protocol.TCAM can only store a small number of flow table entries,which leads to that most packets need to wait for the SDN controller to issue new flow rules for processing,resulting in a large packet forwarding delay.This paper makes an in-depth exploration and optimization of flow table lookups.The research work is as follows:(1)To address the problem of high energy consumption of TCAM flow table lookups in OpenFlow switch,this paper analyzed locality of network traffic,identified flows that frequently arrives in OpenFlow switches as active flows,then cached active and exact flow with low energy consumption SRAM.Most packets can be forwarded directly through the cache,so as to bypass TCAM flow table lookups process and significantly reduce energy consumption of TCAM flow table lookups.Meanwhile,a scalable energy-saving cache structure was designed to adaptively adjust the cache capacity according to dynamic changes of network traffic,and maintain high cache hit rate,so as to achieve flexible energy saving of flow tables.Furthermore,we have derived the cache hit rate and return rate based on exact flow activity model,and proved the monotonicity and concavity of the cache return rate.Then,the cache scaling condition was set to realize real-time perception of network traffic.Finally,this paper has evaluated the performance of the proposed OpenFlow-based flow table energy-saving storage architecture by employing backbone network traffic traces.The experimental results show that the hit rate of the proposed elastic energy-saving cache is always higher than that of the existing energy-saving cache in the case of network traffic fluctuation,and can maintain stability,which significantly improves the energy-saving effect of OpenFlow-based flow table.(2)To address the problem that TCAM can only store a small portion of flow table entries in large-scale network scenarios,which leads to large packet forwarding delay,this paper has proposed a flow table lookup architecture based on flow table aggregation algorithm.We have divided flow table entries into matching domain and content domain,and utilized TCAM and DRAM to store them respectively,so that TCAM can store more flow table entry matching fields.Then,network flows have been divided into wildcard flows and exact flows according to type of flow table entries it matches.Because TCAM supports fast lookups of wildcard entries,it can be utilized to store wildcard flows in network.If number of flows exceeds capacity of TCAM,some flow table entries are split into SRAM.Furthermore,this paper designed an OpenFlow flow table aggregation algorithm,which combines TCAM flow table with the hamming distance of one in the SRAM flow table.Characteristic of the algorithm is that it can solve the problem that the table entries cannot be merged due to the inconsistency of action set by constructing the DRAM action set tree improves the flow table aggregation effect.TCAM can store more flow table entries,which significantly improves search performance of OpenFlow flow table.The experimental results show that most network groups could match the TCAM flow table successfully through OpenFlow flow table aggregation,which effectively improves lookup speed of OpenFlow-based flow table.