|
| Speaker: | Anne Ngu |
| University of New South Wales | |
| Date: | 2nd February 1999 |
| Time: | 11:30am |
| Place: | Seminar Room 357, Building E6A, Macquarie University |
Abstract:
This talk will feature the multibrokering work that I conducted under the InfoSleuth project while on sabbatical at MCC, Austin Texas. InfoSleuth is an agent-based system for information discovery and retrieval in an open and dynamically changing web enviornment. It implements a brokering function which combines reasoning over both the advertised syntax and semantics of agents in the InforSleuth domain.
The brokering function in a distributed or agent-based system is the process of matching processes that require a service (requestors) with processes that can provide a service (service providers). The brokering function must consider two aspects of a potential service provider: (1) whether the potential provider understands the syntax of the request; and (2) whether the potential provider implements the semantics that match the expectations of the requestor. The syntax and semantics describing offered services needs to be defined (advertised) to the broker in terms of a common vocabulary, more complex than that needed just to share syntax. The broker can then reason over this advertised knowledge in order to dynamically determine which agent can provide the requested services. Robustness and scalability considerations dictate that the brokering function must be distributable across a group of collaborating processes. This requires expanding the functionality of individual brokers to allow them to advertise their capabilities to other brokers, as well as receive advertisements from other brokers. Additionally, multibrokering requires the ability of brokers to make requests to other brokers for agent capability information. Our multibroker design is a peer-to-peer system that allows several brokers to collaborate in executing the brokering function when answering a given request. Each broker has the capability to reason over their knowledge, find other brokers, and dynamically determine which of those other brokers they should interact with in the process of matching a request with a service.Enquiries: sals@mri.mq.edu.au
| Last modified: September, 1998 |