Abstract |
This thesis is intended to explore the field of information processing services over computer networks. A modular, flexible and extensible architecture is proposed, that can be used for the development of environments that support information processing services. This architecture is based on a society of distributed autonomous agents that cooperate in order to execute processing algorithms in a distributed environment. This thesis also proposes a distributed resource management mechanism, which is based on the market metaphor and is studied through simulation experiments. The proposed architecture is used for the development of a specific application, an environment for supporting image processing services, DIPE (Distributed Image Processing Environment). DIPE aims to support specialized researchers in the field of image processing as well as common end users, and offers the ability of transparent algorithm execution through the exploitation of computational power and collections of algorithms available in a network. In addition, the proposed environment aims to become the integration platform for various image processing algorithms, either new or developed by third parties (in which case there is no access to their source code). DIPE also offers resource management mechanisms, while quality of service depends on the (virtual) cost the user is prepared to pay for each execution. The ultimate goal of this work is to generalize the design and implementation strategies proposed in DIPE, and lead towards the future development of a framework that can readily support implementation of environments for the provision of information processing services irrespective of the application domain.
|