Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.487513
Title: Transport Control Protocol Optimisation over Wireless Internet: a Cross-Layer Approach
Author: Ewald-Arostegui, Nimbe Leonor
ISNI:       0000 0001 3455 4026
Awarding Body: University of Leeds
Current Institution: University of Leeds
Date of Award: 2008
Availability of Full Text:
Access through EThOS:
Abstract:
This thesis presents an analytical model of a cross-layer communication system to enable improvement in the Transport Control Protocol (rCP) over mixed wired and wireless Internet. The focus is on the quantitative performance evaluation of the interactions between TCP and a hybrid Automatic Repeat reQuest protocol (HARQ) in the link layer (LL) with a finite buffer size. TCP evolution is described and a review of the different approaches to im/ r prove its performance is given. A survey of the most relevant TCP analytical proposals is also provided. The operation of a HARQ scheme comprising Forward Error CorreCtion (FEC) convolutional codes and a Selective Repeat (SR) ARQ protocol is analytically charac~erised by a Discrete Time Markov Chain (DTMC). Corruption losses occur in a radio channel and are modelled by a two-state DTMC whilst congestive losses are due to an LL finite buffer size. HARQ performance parameters are computed through the twO-moment approximation of an M/G/K/l analysis. The accuracy and correctness of the HARQ model is assessed as well as its adequacy for multimedia applications. All analytical results are validated through the well known network simulator, ns-2. It is shown that performance improvemen~ is achieved by dynamically selecting optimal HARQ parameters. It is also illustrated that HARQ can provide the required 'QoS as well as reliable transfer to multimedia applications only under certain power' and traffic conditions given that the impact of congestive losses is greater as the transmission quality improves. An existing mathematical representation of TCP Reno is extended in order to model TCP NewReno given its better performance over noisy channels, Continuous Time Markov Chains (CTMC) are used to this end. The NewReno model accuracy is also validated. .A cross-layer TCP-HARQ communication system is developed with these two analytical models. Top down e>.:plicit notification from TCP to HARQ takes place. The advertised TCP state variables are passed to the LL entity which is able to predict NewReno performance and select the optimal HARQ ,. values. These predictions are computed through fixed-point approximations. The joint TCP-HARQ analytical model is validated and its correctness is demonstrated as well as its capacity to provide a significant TCP performance improvement through the selection of HARQ optimal parameters. It is also demonstrated that smaller LL buffer sizes than the typical size provide similar TCP throughput whilst higher buffer capacities do not represent significant performanc'e improvement. LL buffer capacity reduction mainly depends on the quality of the transmission over the radio channel.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.487513  DOI: Not available
Share: