Research On Active Queue Management Algorithm Of DiffServ-Aware

QoS problem is one of most important researches of computer network area, and researchers are engaged in research on improving QoS indicators. IETF proposed two QoS models include of IntServ and DiffServ, and DiffServ model has been widely used with the advantages of easy expanding, without signaling protocol and so on. There are so many ways to improve QoS indicators, but congestion control is the requirement to ensure QoS. Active queue management is an important method for congestion control, DiffServ-model is also applied widely, so active queue management algorithm combined with DiffServ model has great practical significance.In this thesis, DRIO-C algorithm is proposed after having studied RED and its improved algorithm. The mean idea of DRIO-C algorithm is:first, two functions are added to RED algorithm to make the highest packet loss rates update dynamically in order to make the highest packet loss rates changes with the average queue length value when the average queue length is not between the minimum and maximum threshold; Second, the gradual mechanism of packet loss rates is added when the average queue length is greater than the maximum threshold in order to improve the stability of network when there is a sudden flow; Third, combining with DiffServ strategy in RIO-C algorithm to make DRIO-C support DiffServ. Droping strategy according to different dropping priorities in RIO-C, WRED algorithm is applied widely, but it also has some backwards. In this thesis, MSBLUE algorithm is proposed. The mean idea of MSBLUE algorithm is:First, the minimum and maximum threshold is added to BLUE algorithm and packet loss rates updating formulation is also modified in order to make the average queue length change between the minimum and maximum threshold, keeping the queue length at a reasonable range means that the delay jitter is relatively flat and the stability of the network also has certain guarantee; Second, gradual drop strategy is proposed according to different drop priorities, it increases and decreases packet loss rates of different priority packet in order to completely giving up higher drop priority packet and adequately protecting lower drop priority packet when network resources are relatively insufficient. Two experiments are completed to test DRIO-C and MSBLUE algorithm in NS-2simulation software in this thesis. To test the whole performance and QoS indicator of algorithms, a large network topology which simulates real network is designed in experiment one; to test the effect of DiffServ, it is designed that sending TCP flows with three drop priority in a simple network topology in experiment two. The result from two experiments:At the cost of increasing a little delay, DRIO-C algorithm not only improves the stability of network and throughput but also reduces the packet loss rates. MSBLUE algorithm solves the problem of severely oscillating queue of BLUE, and its DiffServ effect is well, so the performance of low drop priority business will be protected better.