System-Level Issues in Agent Based Distributed Computing (Abstract) (Full paper available from http://www.cs.umn.edu/~tripathi/DartmouthWorkshop.ps)) Anand Tripathi and Neeran Karnik (tripathi|karnik@cs.umn.edu) Department of Computer Science University of Minnesota, Minneapolis, MN 55455 Recently the interest in agent based distributed computing using mobile programs has increased very sharply due the widespread avai- lability and use of the Internet and the large potential for build- ing agent-based distributed computing applications in the Internet domain. The recent development of languages such as Telescript, Java, and Obliq, which support program mobility, is an indicator of the emerging trend in transportable programs. The utility of the agent paradigm in comparison with other widely used paradigms such as message-passing and remote procedure call has been established in the past by other researchers. The primary focus of our work is on the system-architecture and language level issues for agent-based distributed computing. We believe that the advances in distributed object technology over the past 10 years, such as the CORBA model, provide an adequate system architecture basis for agent-based applications. An agent-based application system can be built as a collection of distributed objects, some of which are mobile. This paper identifies the system architecture components for supporting agent-based distributed com- puting. It establishes interrelationships between the elements of an agent-based system, such as "engines", "agent meeting places", and the components of a distributed object management system based on the CORBA model. The paper also identifies the various issues and requirements in supporting the agent paradigm on such an object based distributed system. These requirements are identified based on the needs arising from several different viewpoints related to: object mobility, fault tolerance and status monitoring, client- level control and management of its mobile agents, mechanisms for parallel execution and coordination of agents, multithreading within an agent, and security and protection of information.