Use this URL to cite or link to this record in EThOS:
Title: Forecasting full-path network congestion using one bit signalling
Author: Woldeselassie, M.
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2010
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This dissertation presents a novel approach to Internet congestion control known as Probabilistic Congestion Notification (PCN). Preliminary research has established the requirements and motivations for a more efficient Transmission Control Protocol (TCP) congestion control schemes. Because of the substantial increase in the the number of Internet users and the diversification of online services, the dynamics of the Internet are changing, making classical congestion control mechanisms inefficient. Traditionally, TCP has relied on packet loss as an indication of congestion. Though packet loss is a typical result of overflowing buffers, it cannot be used to prevent congestion before it occurs. Consequently, a novel protocol is proposed in this thesis, which allows congestion notification and control via packet marking. This new protocol is elegant as it makes use of the existing one-bit Explicit Congestion Notification (ECN) field in the Internet Protocol (IP) header. Each PCN data packet can be marked probabilistically by one router at most. Furthermore, PCN performance has been improved by pre-signalling the number of intervening queues along the path. The level of load factor (congestion measure) at each link is fed back to the PCN source, which then estimates the exact level of congestion at each intermediate queue, one-step ahead of time. By knowing this, the source can take avoiding action either by adjusting its sending rate or by using alternate routes. The estimation mechanism uses Multiple Linear Regression (MLR) and Time Series Analysis to improve the quality of the congestion estimate and to predict the level of congestion at each queue along the path. For this purpose, the work presented in this thesis also analyses the suitability and accuracy of such statistical methods in predicting future congestion levels. The new PCN congestion control protocol has been designed and simulated in the network simulator (ns-2). Moreover, this project evaluates PCN’s performance using real high speed Internet traffic traces. Results show that the methods can successfully predict the congestion level at any queue along the path. The proposed approach has low complexity and it is easy to implement. Furthermore, it can be easily deployed in existing TCP/IP networks as it does not require modifications to existing IP or TCP implementations.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available