Please use this identifier to cite or link to this item:
Title: Mobile agent-based routing for the next-generation Internet
Authors: Liu, Wai-tung
Degree: M.Phil.
Issue Date: 2008
Abstract: One of the fundamental issues of the Internet is routing. In general, its purpose is to forward packets over the Internet as efficiently as possible. Currently the Internet is largely based on the well-known TCP/IP protocol suite, in which the transport layer is responsible for the flow control of data communication, while the network layer is responsible for carrying out the routing process. In general, these traditional routing methods, together with the Internet Protocol, have three disadvantages. First, they only offer best-effort delivery service. In other words, no guarantee of support for Quality of Service (QoS). Second, the processing overhead for routing packets is very high since basically each router needs to perform the time-consuming longest prefix match operation. Third, it cannot provide customized routing services. To address the first problem, a number of research projects are being conducted to support QoS in the Internet. These include Differentiated Services (DiffServ), Integrated Services (IntServ) and Resource reSerVation Protocol (RSVP). In addition, a number of QoS routing protocols are emerging to route packets based on QoS requirements. The second problem has been addressed by the introduction of an evolutionary switching technique, known as Multi-Protocol Switching Protocol (MPLS). This technique binds network layer routing to data link layer switching. The third issue, the provision of customized routing services, is being tackled by the proposal of a novel network architecture known as active networks. In this new architecture, packets are active and programmable so as to facilitate the introduction of new or customized services. The active packets function like mobile agents and can thus be run by the network nodes to perform various functions. In recent years, there has been considerable interest in employing mobile agents for various purposes. Initially, mobile agents were mainly used at the application layer but more recently mobile agents have also been employed to address network layer problems like routing. It can be foreseen that mobile agents may be incorporated into the emerging active network framework as a way of providing many innovative services. The aim of this thesis is to explore some of these opportunities. As an extension to previous work, we have studied an active routing service which uses active packets (packlets) to configure customized communication paths based on the network information. With the aim of minimizing network costs, both a finite horizon and an infinite horizon Markov decision model have been formulated to support active routing. We have also studied an active MPLS service with bandwidth reservation using a dynamic pricing mechanism. With the aim of maximizing the profits of the network service provider, an optimal bandwidth selling policy has been obtained for both the finite and infinite horizon models based on the Markov decision theory. In both models the optimal policy is threshold-based with the threshold depending mainly on the amount of remaining bandwidth. It is expected that the Markov decision model will also be adaptable for the analysis of similar bandwidth selling problems in other networking scenarios. Lastly, we have studied a load balancing problem. We have applied a modified network simplex algorithm. This algorithm can employ the traditional network simplex method not only to find multiple MPLS paths for a user application but also balance the load to obtain aggregate minimum network delay for the user application. Active packets (or mobile agents) can be used to implement our modified network simplex algorithm and to set up the relevant label switched paths.
Subjects: Hong Kong Polytechnic University -- Dissertations.
Internetworking (Telecommunication)
Routing (Computer network management)
MPLS standard.
Pages: xii, 93 leaves : ill. ; 30 cm.
Appears in Collections:Thesis

Show full item record

Page views

Last Week
Last month
Citations as of May 28, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.