Performance analysis of TCP connections with RED control and exogenous traffic

Abstract

The Transmission Control Protocol (TCP) is the main protocol used in the Internet for data transfer applications. Certain congestion control mechanisms have been incorporated in host implementations of this protocol which have led to much better network performance. However it seems that besides these congestion control mechanisms one needs buffer management mechanisms like Random Early Detection (RED) in core network routers in order to complement the mechanisms from the edges. RED is also expected to play a m ajor role in the Differentiated Services architecture for providing Quality o f Service (QoS) in the Internet. A typical scenario in tomorrow’s Internet would involve router queues that are shared by traffic emanating both from data transfer applications as well as multimedia applications, which are supposed to be supported on the User Datagram Protocol (UDP). In this thesis we study the stability and performance o f a system involving several TCP connections passing through a tandem of RED controlled queues each of which has an incoming exogenous stream. The exogenous stream which represents the superposition of all other incoming UDP connections into the queue has been modeled as an M M PP stream. We consider both the T C P Tahoe and the TCP Reno versions. We start with the analysis of a single T C P connection sharing a RED controlled queue with an exogenous stream. The effect of exogenous stream (which is almost always present in large networks) is to cause the system to converge to a stationary distribution from any initial conditions. We also obtain the performance indices o f the system; specifically the throughputs and the stationary sojourn times of the various connections. The complexity involved in computation of performance indices for the system is reduced by approximating the evolution of the average queue length process of the RED queue by a deterministic ODE. Then, using a decomposition approachof two time scales, we reduce the study of the system to that of a simplified one for which performance measures can be obtained under (quasi)stationarity. Finally we extend the above results to the case when multiple TCP connections share a RED controlled queue with an exogenous stream and to the case when T C P connections pass through several RED controlled routers each of which has an incoming exogenous stream. A number of simulation results have been provided which support the analysis.