Tuncer, DS; (2013) Engineering self-managed adaptive networks. Doctoral thesis , UCL (University College London).

Preview

PDF PhDThesis_DaphneTuncer.pdf Available under License : See the attached licence file. Download (6MB)

Abstract

In order to meet the requirements of emerging services, the future Internet will need to be flexible, reactive and adaptive with respect to arising network conditions. Network management functionality is essential in providing dynamic reactiveness and adaptability but current management approaches have limitations which prevent them from meeting these requirements. In search for a paradigm shift, recent research efforts have been focusing on autonomic/self-management principles, whereby network elements can adapt themselves to contextual changes without any external intervention through adaptive and flexible functionality. This thesis investigates how autonomic principles can be extended and applied to fixed networks for quality of service and performance management. It presents a novel resource management framework which enables intelligence to be introduced within the network in order to support self-management functionality in a coordinated and controllable manner. The proposed framework relies on a distributed infrastructure, called the management substrate, which is a logical structure formed by the ingress nodes of the network. The role of the substrate is illustrated on realistic resource management application scenarios for the emerging self-managed Internet. These cover solutions for dynamic traffic engineering (load balancing across multiple paths), energy efficiency and cache management in Internet Service Providers. The thesis addresses important research challenges associated with the proposed framework, such as the design of specific organisational, communication and coordination models required to support the different management control loops. Furthermore, it develops, for each application scenario, specific mechanisms to realise the relevant resource management functionality. It also considers issues related to the coexistence of multiple control loops and investigates an approach by which their interactions can be managed. In order to demonstrate the benefits of the proposed resource management solution, an extensive performance evaluation of the different mechanisms described in this thesis have been performed based on realistic traffic traces and network topologies.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item