Multiple TCP Connections Research Articles

Stability and Analysis of TCP Connections with RED Control and Exogenous Traffic

In this paper we study the stability and performance of 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, representing the superposition of all incoming UDP connections into a queue, has been modeled as an MMPP stream. We consider both the TCP Tahoe and the TCP Reno versions. We start with the analysis of a single TCP connection sharing a RED controlled queue with an exogenous stream. The effect of the 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. This stability result holds good for any work conserving discipline. We also obtain the performance indices of the system; specifically the goodputs and the mean 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, by using a decomposition approach of two time scales, we reduce the study of the system to that of a simplified one for which the performance measures can be obtained under 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 a TCP connection passes through several RED controlled tandem queues each of which has an incoming exogenous stream. We also consider an example of multiple TCPs passing through a tandem of queues. A number of simulation results have been provided which support the analysis.

Using bundles for Web content delivery

The protocols used by the majority of Web transactions are HTTP/1.0 and HTTP/1.1. HTTP/1.0 is typically used with multiple concurrent connections between client and server during the process of Web page retrieval. This approach is inefficient because of the overhead of setting up and tearing down many TCP connections and because of the load imposed on servers and routers. HTTP/1.1 attempts to solve these problems through the use of persistent connections and pipelined requests, but there is inconsistent support for persistent connections, particularly with pipelining, from Web servers, user agents, and intermediaries. In addition, the use of persistent connections in HTTP/1.1 creates the problem of non-deterministic connection duration. Web browsers continue to open multiple concurrent TCP connections to the same server. This paper examines the idea of packaging the set of objects embedded on a Web page into a single bundle object for retrieval by clients. Based on measurements from popular Web sites and an implementation of the bundle mechanism, we show that if embedded objects on a Web page are delivered to clients as a single bundle, the response time experienced by clients is better than that provided by currently deployed mechanisms. Our results indicate that the use of bundles provides shorter overall download times and reduced average object delays as compared to HTTP/1.0 and HTTP/1.1. This approach also reduces the load on the network and servers. Implementation of the mechanism requires no changes to the HTTP protocol.

An accurate model for analyzing wireless TCP performance with the coexistence of Non-TCP traffic

Our work reported in this paper attempts to develop a detailed analytical model of the TCP congestion control algorithm in wide-cum-wireless environments with regular TCP traffic that could coexist with other exogenous connections such as voice and video traffic. Thus our model is unlike previous models that did not explicitly consider the impact of non-TCP traffic on the referenced TCP connection. We think that it is extremely difficult and also unnecessary to try to use a single closed-form model to accurately describe the TCP performance. Based on an approximated fluid model, we deduce the TCP throughput performance with arbitrary connection size and the coexistence of non-TCP traffic. Most previous literature on TCP models focused on steady-state TCP congestion control behavior. However, a wealth of evidence suggests that many popular TCP connections are short, often below 30 KB. This paper examines how well our proposed model could capture the behavior of short TCP connections with wireless error losses. The simulation results show that our hybrid method approximates the throughput performance of an arbitrary-sized TCP connection with wireless losses much better than other proposed models. In addition, the throughput performance of multiple TCP connections is analyzed and simulated. Based on our analysis results, we make some suggestions for TCP performance enhancement for overcoming the serious effect from wireless losses.

Predicting bottleneck bandwidth sharing by generalized TCP flows

The paper presents a technique for computing the individual throughputs and the average queue occupancy when multiple TCP connections share a single bottleneck buffer. The bottleneck buffer is assumed to perform congestion feedback via randomized packet marking or drops. We first present a fixed point-based analytical technique to compute the mean congestion window sizes, the mean queue occupancy and the individual throughputs when the TCP flows perform idealized congestion avoidance. We subsequently extend the technique to analyze the case where TCP flows perform generalized congestion avoidance and demonstrate the use of this technique under the Assured Service model, where each flow is assured a minimum traffic rate. Simulations are used to demonstrate the accuracy of this technique for relatively low values of packet dropping probability and a much wider range of packet marking probability.

A mean-field model for multiple TCP connections through a buffer implementing RED

Active queue management schemes like RED (random early detection) have been suggested when multiple TCP sessions are multiplexed through a bottleneck buffer. The idea is to detect congestion before the buffer overflows and packets are lost. When the queue length reaches a certain threshold RED schemes drop/mark incoming packets with a probability that increases as the queue size increases. The objectives are an equitable distribution of packet loss, reduced delay and delay variation and improved network utilization. Here we model multiple connections maintained in the congestion avoidance regime by the RED mechanism. The window sizes of each TCP session evolve like independent dynamical systems coupled by the queue length at the buffer. We introduce a mean-field approximation to one such RED system as the number of flows tends to infinity. The deterministic limiting system is described by a transport equation. The numerical solution of the limiting system is found to provide a good description of the evolution of the distribution of the window sizes, the average queue size, the average loss rate per connection and the total throughput. TCP with RED or tail-drop may exhibit limit cycles and this causes unnecessary packet delay variation and variable loss rates. The root cause of these limit cycles is the hysteresis due to the round trip time delay in reacting to a packet loss.