Research On Traffic Scheduling Algorithm In Time Sensitive Networks

Time Sensitive Network(TSN)is a new generation of Ethernet switching technology that can provide deterministic transmission guarantee.In TSN,the high-priority scheduled traffic can be carried by scheduled flows(SF).The transmission is controlled by the gate mechanism of IEEE 802.1Qbv.Given the network and SF information of a TSN network,the SF scheduling problem is calculating all the Gate Control Lists(GCL)of egress ports of the network.For the requirement of practical application of TSN,this thesis focuses on the above traffic scheduling problems.First,this thesis models the TSN traffic scheduling problem.Combining the actual scenario of TSN and the existing methods,the Time Sensitive Network scheduling problem is mapped to the No-wait shop scheduling problem.On this basis,a unified mathematical model is established.The obtained optimization problem is NP-hard.In this thesis,by linearizing the constraints of the problem,an integer linear programming(ILP)problem is formed,which can be solved by ILP solving tools.And the limitations of ILP solution are shown through simulation and analysis.Then,the scheduling algorithm in dynamic scenario is studied.In a dynamic network,the traffic load is light and requires fast calculation.As the number of possible conflicting transmissions in the network increases,it takes a long time to solve the problem using ILP solvers,which is no longer suitable for dynamic scenarios.This thesis proposes a PHS-Based Greedy Scheduling Algorithm(PHSGSA).The simulation results in different scenarios show that PHSGSA can quickly and effectively solve the TSN scheduling problem in light load scenarios.Finally,this thesis studies the scheduling algorithm of large-scale static scenarios.In the static scenarios,it is allowed to produce higher quality scheduling results through longer calculation-time.Based on the established mathematical model,this thesis proposes a scheduling result evaluation criterion that conforms to static scenarios to achieve multi-objective optimization of network utilization and remaining bandwidth.On this basis,a Flexible Mixed initial population Genetic Algorithm(FMGA)is further proposed.FMGA can adaptively generate a mixed initial population with better performance according to the problem characteristics in different scenarios.A simulation program is designed to compare the scheduling performance of various algorithms in different settings such as network node size,SF period characteristics,network traffic load,etc.The results show that compared with existing algorithms,FMGA has lower complexity and can obtain better scheduling results in most simulation scenarios.