1 Department of Photonics Engineering, Technical University of Denmark2 Networks, Department of Photonics Engineering, Technical University of Denmark3 Department of Electromagnetic Systems, Technical University of Denmark4 Copenhagen Center for Health Technology, Center, Technical University of Denmark
This thesis - Resource Management in Broadband Communication Networks - deals with different ways of optimizing the available resources of data- or telecommunication networks. Especially topics like optimal routing, load balancing and fast recovery of routes in case of link failures are covered. The first part gives a brief description of some of the existing protocols for routing and controlling resources, such as RSVP, OSPF, BGP, PNNI, etc. The remaining part concerns the following two fundamentally different approaches to resource management etc.: - Software agents - Simulated ants In the beginning of the part concerning software agents a description of what exactly constitutes software agents, according to the scientists and according to the organization, FIPA is given. After this, the main results from the IMPACT project are presented. The IMPACT project is an EU-project that aimed at developing a demonstration platform, where software agents handled virtually all aspects of controlling an ATM-based network. In the beginning of the part about simulated ants it is explained how scientists believe ants in nature find their way through the environment between their nest and food sources. These observations are converted into a model, which in turn leads to a Java-based ant-simulator, that can be used to thoroughly inspect the behaviour of the simulated ants. Furthermore, the model is changed in several ways, which means that the simulator is capable of investigating 10 different ant classes if some of them with improved characteristics, and others with worse characteristics (compared to the original model). The results from the investigation of the use of software agents in the context of networks is primarily qualitative, since it was not possible to make a lot of measurements that could in reality be compared to existing solutions. Instead, it was possible with the implemented software agent to demonstrate scenarios, where the agents could fit in additional connections by negotiations and borrowing bandwidth from each other. In addition, it was considered as an important part to build a model, that reflects the situation in a competitive telecommunication market, with several service providers sharing the same physical network. The investigation of the simulated ants, on the other hand, gives a many promising results, suggesting that they can successfully be used to solve some problems in communication networks. For instance, the results show that a network controlled by simulated ants can balance the load quickly and efficiently, thereby postponing local hot-spots or in some cases even avoid hot-spots. Furthermore, the results clearly demonstrate, that systems using simulated ants obtain a virtually unprecedented robustness. A network with failing components typically return to a fully operational state within seconds or even faster. Hopefully these results of the investigation of simulated ants - along with other results from the literature - can contribute to make the big manufacturers of telecommunication equipment consider using simulated ants in their future products.