This thesis is devoted to Optical Burst Switched (OBS) networks, being focused on presenting new solutions to OBS network performance as a whole, from the ingress to the core nodes including the burst assembly, switching and routing tasks.
We present several new solutions to problems of the OBS networks, namely, a new burst assembly machine concept named Internet Protocol Packet Aggregator and Converter (IP-PAC), with a new burst assembly algorithm, dynamic and adaptive to network traffic fluctuations. Furthermore we propose the use of this machine not only in OBS networks but also as an aggregation device for other networks. We identify several benefits from the use of the IP-PAC concept, in particular its contribution in eliminating bottleneck problems.
We also present two new routing algorithms, Extended Dijkstra and Next Available Neighbour. The first is a balance and symmetry concerned algorithm that keeps the features of the Dijkstra Algorithm, i.e. it remains a shortest path simple algorithm. It is suitable to use in simulation as its behaviour tends to be closer to the behaviour of real networks since it does not overload unnecessarily some links more than others. It may also be used in machines where the computation of paths is Dijkstra based and situations of equal-cost routes may exist. Its performance is evaluated for OBS networks. The second algorithm is a dynamic non-deterministic routing algorithm that is applicable to OBS and to other networks. Its main feature is that in a situation of imminent burst loss, the burst is routed to another node, being this neighbour node the first available according to a series of metrics. As these metrics are used in a particular undetermined moment of time, the result is non-deterministic routing.
The study of the characteristics of burst traffic is of capital importance to understand the behaviour of OBS networks. As a pre-requisite to this aspect of the research, we study OBS networks tributary IP traffic, using a series of recorded real IPv4 packet traces. In this thesis, we conclude that the main OBS network performance metric, burst loss ratio, is equivalent to other metrics like packet loss ratio and byte loss ratio when burst assembly is performed with efficiency concerned algorithms and using real traffic.
We assess the efficiency of main burst assembly algorithms and propose a new dynamic burst assembly algorithm, with thresholds that adapt to traffic conditions to allow an optimized burst assembly process. While assessing the assembly of bursts, we found that in real traffic conditions most bursts will be at around 9 KB of size. We also study the effect of the implementation of larger IP packets size for IPv6 and conclude that there are routing and switching benefits to reap from the usage of larger IP packets. Furthermore, we conclude that for burst assembly tasks, IPv4 and IPv6 behave similarly and thus the conclusions drawn on IPv4 datagrams can be extended to IPv6 packets, including Jumbograms.
The main contribution of this thesis is the proposal of a new OBS architecture, named Common Control Channel OBS, or C3-OBS for short. In this architecture we propose the passive broadcast of the control packets in a special tree-like control channel, as a way to disseminate the information throughout the network. We then propose the use of a Local Network Model (LNM) database structure at each node to allow concise network management and behaviour prediction. In the C3-OBS nodes we propose and test a new routing and scheduling algorithm we named Travel Agency Algorithm. We analyse some problems that rise in this new approach and propose solutions, a new approach using network domains for OBS as a way to minimize the flooding of the network with control packets, and some special features on the Travel Agency Algorithm as a way to identify and solve concurrent reservation situations. We assess and compare the performance of the C3-OBS architecture with regular OBS architecture for several topologies.
Finally we present the research conclusions and propose directions for future work.