Active queue self-adaptive management method ASRED

Abstract

The invention provides an active queue self-adaptive management method ASRED. Active queue management AQM has always been a hot issue in the research of Internet congestion control. Although the traditional AQM algorithm can effectively predict and control the occurrence of network congestion, however, the traditional AQM algorithm relies on parameter setting too much and the sensitiveness is high, in different networks, the algorithm performance is unsteady and the performance difference is greater. Aiming at the above problems, a new active queue management algorithm ASRED is designed on the basis of GRED (Gentle RED) and ARED (Adaptive RED) algorithm thoughts, and a nonlinear function is adopted to calculate the probability of packet loss, and the quantity of packet loss is controlled more reasonably in a smooth curve growth mode; meanwhile, due to the characteristic of self-adaptive adjustment of max, the active queue self-adaptive management method ASRED provided by the invention can better adapt to the changes of the network environment, predict and prevent the occurrence of congestion, and can adequately guarantee the network throughput rate for the changeable network environment.

Description

technical field [0001] The invention belongs to the technical field of Internet. Background technique [0002] In recent years, with the rapid development of the network, simply using the source-side congestion control mechanism can no longer meet the user's increasingly high requirements for network performance. The congestion control algorithm of intermediate nodes opens up a new research direction for the research of congestion control. The active queue management algorithm plays an extremely important role in the congestion control mechanism of intermediate nodes and is widely used in practice. Among them, the RED algorithm is the most representative, but the performance of the algorithm depends too much on the parameter setting, and the algorithm parameters are sensitive High degree, stability and robustness still have room for improvement, the following is an analysis of the advantages and disadvantages of several classic active queue management algorithms: [0003] ...

AI Technical Summary

Problems solved by technology

[0038] (1) The algorithm does not solve the problem of parameter sensitivity, and its performance still depends on the parameter settings, and different network environments still require different parameters

[0039] (2) Simply changing the growth rate of the packet loss probability cannot achieve real-time adjustment of the growth rate, and the smoothness needs to be improved

[0040] (3) If the queue length has been increasing, the average queue length QL avg Equal to the maximum domain max th , the algorithm still faces the problem of "global synchronization"

Images