Congestion-Minimized Routing Update Strategy In Software-Defined Networking

Software-Defined Networking(SDN)centralizes management and control functions in the controller.In order to achieve goals such as high link utilization,the SDN controller needs to frequently change the transmission path of the data flow.However,during the routing update process,the links in the network may be congested transiently.Congestion will cause packet loss and retransmission,reduce the original normal sending rate,and increase transmission delay.Therefore,designing a routing update strategy with the least packet loss has become a research hotspot.The most ideal update strategy can avoid congestion.However,the congestion-free update process takes a long time,which limits the frequency of performance tuning by the controller,and prevents the controller from responding to unexpected situations such as link failures in time.Therefore,under the premise of avoiding congestion,reducing the update time is also of great significance.Existing research uses optimization tools to find congestion-minimized routing update strategy.If congestion is inevitable,in order to minimize the amount of packet loss,the existing optimization model uses link utilization to measure the amount of packet loss,and the designed update strategy will still produce many situations where the amount of packet loss varies greatly.In order to avoid congestion in the update process,in addition to using optimization tools,existing research also proposes a routing update algorithm based on dependencies.However,the current algorithm only migrates the data flow once,and a longer dependency chain will be formed between migrations,resulting in longer update time.In view of the congested routing update process,this paper constructs an optimization model based on rate limitation,and measures the packet loss by the numerical value of limited rate.Since solving the model is NP-Hard,this paper designs a two-step heuristic algorithm MIC(Minimizing the Impairment of Congestion).The first step of MIC is to determine the intermediate state,limit the number of congested transitions,and establish the state transition sequence.The second step of MIC is to minimize the limited traffic for each congested transition process.The simulation experiment results show that compared with the algorithm based on link utilization,MIC can limit the number of congested state transitions and reduce the amount of packet loss by 89%.It is worth noting that MIC can distinguish the importance of data flows and reduce the packet loss of data flows with high importance.In order to reduce the update time under the premise of no congestion,this paper designs the Twin(Two migrations),a routing update algorithm based on two migration.Twin first segment the data flows.Based on the segmented data flows,Twin constructs an update sequence,and migrates the data flows in the middle of the update sequence or the deadlocked data flows twice,that is,first migrate to the path with free bandwidth resources,and then migrate to the destination path.Due to the time-consuming jitter of path migration,in order to start the path migration of two migration data flow as soon as possible,Twin designs a path selection mechanism during the update process.Simulation experiments prove that Twin can reduce the average operation time by 9.8% and the required computation time is shorter compared to the exiting algorithm based on dependencies.It is worth noting that when the effect of the segmentation method is weakened,the performance decline of Twin is smaller than that of the existing algorithm,indicating that Twin has a wider application range.MIC can reduce the amount of packet loss caused by link congestion and avoid unnecessary jitter caused by route update.At the same time,MIC can reduce packet retransmission,reduce transmission delay,and avoid the loss of delay sensitive applications caused by route update.Twin can reduce the update time of route update process without congestion,support SDN controller to carry out frequent network management operation,and make SDN controller respond quickly to unexpected circumstances such as link failure.