| The future avionic network is an integrated communications network with security,real-time,reliability,opening-up and standardization characteristics.As a new real-time network technique based on time-triggered protocol,TTE network is compatible with the traditional Ethernet and AFDX airborne network,supporting mixed-critical traffics transmission,meeting the requirements of future avionics system and is widely known as an ideal avionics network technology.The switching and scheduling of real-time traffics are the key aspects for system implementation,which directly affect the overall performance of the avionics system.The thesis focuses on the static schedule design in different scheduling methods,and the performance evaluation of real-time traffics.To reduce the waiting latency of RC traffics,we propose a partition scheduling algorithm(RM-LC)based on Rate-Monotonic and the Left-Compression principle.We analyze the TTE network performance using network calculus,and derive the theoretical upper latency bounds of TT and RC single-traffic.We reduce the pessimism of the traditional theoretical calculation through applying Pay Burst Only Once Principle,and obtain the theoretical latency of RCVL in FIFS and SPS queuing strategies respectively.The feasibility of RM-LC and the correctness of related theoretical analysis are verified in OPNET simulation platform.Compared with the general RMS algorithm,RM-LC provides the performance improvements in Latency Mean and Maximum Latency of all the RCVLs.SPS assures the real-time,and reduces delay-jitter of high-priority RCVLs by sacrificing the latency performance of low-priority RCVLs.In order to solve the starvation problem caused by scheduling TT back-to-back frames,some scholars propose a concept of Porosity Scheduling,but there are no concrete implementation algorithms and related theoretical analysis on the method.We present a porosity scheduling algorithm based on maximizing the number of porosity(MNP),analyze the porosity theoretical latency,and verify the feasibility of MNP by simulation.In order to compare the performance of Partition and Porosity,we construct the Switch Latency Models of RC aggregate flow in the two scheduling methods using network calculus.We consider the Latency Mean of all the RCVLs as the evaluation index,given the condition that all TTs are schedulable,and deal with the performance difference in various parameters of the two scheduling methods.Finally,the validation is made via simulations.The results show that the relative merits of Partition and Porosity with GI will be changed when Guard Interval or Burst Degree changes,but the latency performance of Porosity without GI is always optimal. |
