Year

2006

Season

Spring

Paper Type

Master's Thesis

College

College of Computing, Engineering & Construction

Degree Name

Master of Science in Computer and Information Sciences (MS)

Department

Computing

First Advisor

Dr. Sanjay Ahuja

Second Advisor

Dr. Zornitza Prodanoff

Third Advisor

Dr. Robert Roggio

Department Chair

Dr. Judith Solano

College Dean

Dr. Neal Coulter

Abstract

As high-speed networks with large bandwidth delay products (BDP) become more common, high-speed transport protocols must be developed that perform well in these contexts. TCP has limitations in high BDP networks. A number of high-speed TCP proposals have emerged, including BIC TCP, High Speed TCP, and H-TCP. XCP is an intraprotocol communication mechanism that promises even greater performance by providing explicit feedback from routers about congestion. It requires changes to routers and end hosts, though, whereas the other experimental protocols only require changes to an end host. We evaluated the performance ofXCP against BIC TCP, High Speed TCP, H-TCP, and . NewReno TCP. We found that in a controlled environment, XCP gave much better performance than the other TCPs. XCP was sensitive to misconfiguration and environmental factors, though, and was more difficult to deploy. More work is required to make XCP more stable. The other TCPs did not perform better than NewReno TCP but show promise, as most performed almost as well as NewReno TCP.

Share

COinS