Use this URL to cite or link to this record in EThOS:
Title: Towards automatic traffic classification and estimation for available bandwidth in IP networks
Author: Lai, Zhaohong
ISNI:       0000 0004 2673 4205
Awarding Body: University of London
Current Institution: University College London (University of London)
Date of Award: 2008
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Growing rapidly, today's Internet is becoming more difficult to manage. A good understanding of what kind of network traffic classes are consuming network resource as well as how much network resource is available is important for many management tasks like QoS provisioning and traffic engineering. In the light of these objectives, two measurement mechanisms have been explored in this thesis. This thesis explores a new type of traffic classification scheme with automatic and accurate identification capability. First of all, the novel concept of IP flow profile, a unique identifier to the associated traffic class, has been proposed and the relevant model using five IP header based contexts has been presented. Then, this thesis shows that the key statistical features of each context, in the IP flow profile, follows a Gaussian distribution and explores how to use Kohonen Neural Network (KNN) for the purpose of automatically producing IP flow profile map. In order to improve the classification accuracy, this thesis investigates and evaluates the use of PCA for feature selection, which enables the produced patterns to be as tight as possible since tight patterns lead to less overlaps among patterns. In addition, the use of Linear Discriminant Analysis and alternative KNN maps has been investigated as to deal with the overlap issue between produced patterns. The entirety of this process represents a novel addition to the quest for automatic traffic classification in IP networks. This thesis also develops a fast available bandwidth measurement scheme. It firstly addresses the dynamic problem for the one way delay (OWD) trend detection. To deal with this issue, a novel model - asymptotic OWD Comparison (AOC) model for the OWD trend detection has been proposed. Then, three statistical metrics SOT (Sum of Trend), PTC (Positive Trend Checking) and CTC (Complete Trend Comparison) have been proposed to develop the AOC algorithms. To validate the proposed AOC model, an avail-bw estimation tool called Pathpair has been developed and evaluated in the Planetlah environment.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available