This Is AuburnElectronic Theses and Dissertations

Performance Evaluation of Biased Queue Management

Date

2006-08-15

Author

Li, Xiaoming

Type of Degree

Thesis

Department

Computer Science and Software Engineering

Abstract

Congestion is an important issue which researchers focus on in the Transmission Control Protocol (\textit{TCP}) network environment. To keep the stability of the whole network, congestion control algorithms have been extensively studied. Queue management method employed by the routers is one of the important issues in the congestion control study. Biased Queue Management (\textit{BQM}) is a queue management method proposed in [1], and consists of an accurate packet loss discriminator and an implicit continuous congestion level measure. We can apply BQM in both wired and wireless circumstance. Adding the accurate packet loss discriminator into BQM, we can easily distinguish congestion losses and wireless random losses ( \textit{we refer to any loss unrelated to congestion as wireless loss}), and determine when to adjust the congestion window size to slow down the output. Furthermore, with continuous congestion level measure, BQM can relieve the congestion and improve the performance of queue method in TCP environment. Moreover, we will propose a new scheme (\textit{Modified BQM}) for the accurate loss discriminator to improve BQM in some respects of congestion control in ad hoc network. In order to show the improvement achieved by BQM and modified BQM, we will compare BQM with a popular queue management method, Droptail, in different aspects, such as throughput, packet loss rate and fairness. We observe the performance of BQM in both wired and wireless TCP network environment. The comparison results indicate BQM has better throughput, higher fairness and lower packet loss rate than Droptail. BQM is a highly efficient queue management technique in congested TCP network environment, while modified BQM can get better throughput than normal BQM. Experiments are performed by Network Simulator (\textit{NS}) from Lawrence Berkeley Labs with wireless extension from CMU.