Toward a System for Mobile Agent Optimization

 

The performance benefits derived from agent morphing and fusion presented in Sections  2 and  3 are significant. They are motivating us to construct an agent system within which agent/object-objent-programs are easily constructed and adapted at runtime. Such a system consists of contracts for performance requirement specifications, a notification system for contract monitoring, policies for application specified adaptation enactment, and finally, system or application defined adaptations.

This paper only addresses adaptation methods and adaptation enactment mechanisms unique to mobile agent-based applications. In our ongoing research, we are identifying other aspects unique to mobile agents and therefore, appropriately addressed by an adaptive Objent system. We expect to base this work on previous and current research in distributed adaptive systems[34, 6] and in distributed object systems[45].

Specifically, our Objent system has to address the following issues to support morphing and fusion adaptation:

1. Where and how are native agent forms created and maintained?
2. How does the system ensure consistency between the migratory and native versions of agent state?
3. When an agent's form is being or has been changed, can external agents not aware of this change continue invoking it?
4. What are useful fusion algorithms?

The basic fundamental components of the `Objent' system we are developing have been described in a previous publication[2]. We next outline our solution approaches to the specific questions posed above.

• Acquiring an Agent's Native Version.
• Consistency Between Multiple Agent Forms.
• Agent Invocation.
• Fusion Algorithms.