Digital Libraries (DL) generally contain a collection of independently maintained data sets, in different formats, which may be queried by geographically dispersed users. The general problem of managing such large digital data archives is particularly challenging when the system must cope with data which is processed on demand. This dissertation proposes a Multi-Agent System (MAS) architecture for the utilisation of an active DL that provides computing services in addition to data-retrieval services, so that users can initiate computing jobs on remote supercomputers for processing, mining, and filtering of the data in the library. The system architecture is based on a collaborative set of agents, where each agent undertakes a pre-defined role, and is responsible for offering a particular type of service. The integration of services is based on a user defined query which can range in complexity from simple queries, to specialised algorithms which are transmitted to image processing archives as mobile agents. The proposed architecture enables new information sources and services to be integrated into the system dynamically, supports autonomous and dynamic on-demand data processing based on collaboration between agents, capable of handling a large number of concurrent users. Focus is based on the management of mobile agents which roam through the servers that constitute the DL to serve user queries. A new load balancing scheme is proposed for managing agent load among the available servers, based on the system state information and predictions about lifetime of agent tasks and server status. The system architecture is further extended by defining a gateway to provide interoperability with other heterogeneous agent-based systems. Interoperability in this sense enables agents from different types of platforms to communicate between themselves and use services provided by other systems. The novelty of the proposed gateway approach lies in the ability to adapt an existing legacy system for use with the agent-based approach (and one that adheres to FIPA standards). A prototype has been developed as a proof-of-concept to outline the principles and ideas involved, with reference to the Synthetic Aperture Radar Atlas (SARA) DL composed of multi-spectral remote-sensing imagery of the Earth. Although, the work presented in this dissertation has been evaluated in the context of SARA DL, the proposed techniques suggest useful guidelines that may be employed by other active archival systems.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:583550 |
| Date | January 2005 |
| Creators | Georgousopoulos, C. |
| Publisher | Cardiff University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://orca.cf.ac.uk/55153/ |
Page generated in 0.0024 seconds