Enhancing Window-Based Congestion Control Algorithm for the Internet

No Thumbnail Available

Files

Matthew Ayemowa Pre Pages.pdf (906.99 KB)
Matthew AYEMOWA Thesis.pdf (8.07 MB)

Date

2022-12

Authors

Journal Title

Journal ISSN

Volume Title

Publisher

Lead City University

Abstract

The Transmission Control Protocol (TCP) or Internet Protocol (IP) has gone through a number of changes. There have been numerous ideas to alter the mechanisms of TCP congestion control in order to improve its performance. Many research had been done on congestion avoidance and control; however, previous studies on window-based congestion control were mostly based on the starting network, adaptive congestion, and end-to-end mechanisms; however, network overloading will be avoided, and packet losses should be minimized by using the PERC Algorithm. The goal of this study is to develop a Proactive Explicit Rate Control (PERC) algorithm for max-min fair rates that avoids network overcrowding and packet losses through sharing. For max-min fair rates, we devised the PERC algorithm, which can prioritize short flows as needed. Motivated by the need for fast congestion control, this thesis focuses on a different class of congestion control algorithms, called proactive explicit rate-control (PERC) algorithms, which decouple the rate calculation from congestion signals in the network. The switches and Network Interface Cards(NICs) exchange control messages to run a distributed algorithm to pick explicit rates for each flow. PERC algorithms proactively schedule flows to be sent at certain explicit rates. They take as input the set of flows and the network link speeds and topology, but not a congestion signal. As a result, they converge faster and their convergence time depends only on fundamental” dependency chains,” essentially couplings between links that carry common flows, that are a property of the traffic matrix and the network topology. We argue that congestion control should converge in a time limited only by fundamental dependency Our main contribution is to use PERC Algorithm to control Network Overloading and to equally control packet loss In simulation and on a P4 programmed FPGA hardware test bed, s-PERC converges an order of magnitude faster than reactive schemes such as TCP, DCTCP, and RCP. Keywords: Congestion Control, Congestion Window, Network Overloading, Network Protocols, Packet Losses. Word Count: 300

Description

Keywords

Citation

Kate Turabian

URI

https://repository.lcu.edu.ng/handle/123456789/270

Collections

Department of Computer Science;