vioft2nntf2t|tblJournal|Abstract_paper|0xf4ff514b0b0000006f4c010001000800 There are many factors that are important in the design of queue management schemes for routers in the Internet: for example, queuing delay, link utilization, packet loss, energy consumption and the impact of router buffer size. By considering a fluid model for the congestion avoidance phase of Additive Increase Multiplicative Decrease (AIMD) TCP, in a small buffer regime, we argue that stability should also be a desirable feature for network performance. The queue management schemes we study are Drop Tail and Random Early Detection (RED). For Drop Tail, the analytical arguments are based on local stability and bifurcation theory. As the buffer size acts as a bifurcation parameter, variations in it can readily lead to the emergence of limit cycles. We then present NS2 simulations to study the effect of changing buffer size on queue dynamics, utilization, window size and packet loss for three different flow scenarios. The simulations corroborate the analysis which highlights that performance is coupled with the notion of stability. Our work suggests that, in a small buffer regime, a simple Drop Tail queue management serves to enhance stability and appears preferable to the much studied RED scheme.
Ganesh Patil1, Sally McClean2 and Gaurav Raina3
1Indian Institute of Technology Madras, India,2University of Ulster, Ireland,3Indian Institute of Technology Madras, India
TCP, Queue Management, Small Buffers, Performance
| January | February | March | April | May | June | July | August | September | October | November | December |
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Published By :
ICTACT
Published In :
ICTACT Journal on Communication Technology
( Volume: 2 , Issue: 2 , Pages: 339-344 )
Date of Publication :
June 2011
Page Views :
140
Full Text Views :
|