Research On Segment Routing List Generate Algorithm For Network Latency

Segment Routing(SR)is a network layer traffic engineering idea,the core of which is to replace the hop-by-hop destination label in the source route with a segmented routing destination,this path coding approach can be applied to traffic engineering,performance evaluation,failure recovery,and other directions.As network services place increasingly high demands on latency,the network needs to consider the effects of latency guarantees at all levels,and SR which is born at the network layer is ideally suited for latency guarantees.SR has been widely used in network traffic engineering.The SR traffic engineering uses SR policies to guide the forwarding of packets.The policy is expressed as a list of SR transpoints in the packet header,which means any policy of SR will eventually be implemented as a transpoint list.However,the current research objectives of SR transpoint list generation algorithms are often focused on optimizing bandwidth usage efficiency or improving network throughput,resulting in a lack of latency guarantee.To address this problem,this paper proposes two latency-guaranteed SR transpoint list generation algorithms based on the architecture of SR,from the perspective of centralized control and distributed control respectively.Firstly,the first SR transpoint generation algorithm is proposed based on the centralized control of SR.The algorithm establishes link weights for each link considering queuing latency through the global perspective of softwaredefined networking,further reduces the dimensionality of the data model of the network topology through auxiliary graphs based on topology abstraction,and applies the Bellman-Ford algorithm to calculate the SR transpoint list on the reduced data model.The whole algorithm is optimized with the main objectives of reducing the end-to-end latency of SR traffic and reducing the time complexity of the SR transpoint list generation algorithm and finally arrives at an SR transpoint list with a latency guarantee effect.Then,this paper proposes a second SR transpoint generation algorithm based on the distributed control of SR.The algorithm firstly designs an adjacent SR node discovery algorithm for planning the object of SR node latency detection.Secondly,the algorithm generates a latency matrix at each SR node to record the target latency of traffic engineering and designs a latency matrix update algorithm to ensure that the latency information can be obtained by every SR node in a distributed way and updated in its latency matrix.Finally,it designs an SR transpoint list generation algorithm so that the entry SR node can independently compute an SR transpoint list with the specified latency guarantee effect.Finally,this paper implements the above two algorithms based on P4 and BMv2 and performs network experiments to verify them.The control group of the centralized SR transpoint generation algorithm is set to the algorithm that only uses bandwidth to calculate the SR transpoint list,and it is experimentally verified that the latency generated by scheduling traffic using the centralized algorithm proposed in this paper is 8.22%and 3.34%lower than that of the control group in the data center network and the ISP network respectively.The control group of the distributed SR transpoint generation algorithm is set to the shortest path algorithm,and it is experimentally verified that the latency generated by scheduling traffic with the distributed algorithm proposed in this paper in the data center network and the ISP network is reduced by 53.3%and 65.6%,respectively,compared with the control group under the loss of certain network throughput.The experimental results show that the calculation of SR transpoint lists based on latency information can reduce traffic end-to-end latency and provide network technical support for latency-sensitive services.