The Stability, Scalability and Performance of Multi-agent Systems
BT Technology Journal
Distributed Dynamic Capacity Contracting: A Congestion Pricing Framework for Diff-Serv
MMNS '02 Proceedings of the 5th IFIP/IEEE International Conference on Management of Multimedia Networks and Services: Management of Multimedia on the Internet
A Distributed Asynchronous Execution Semantics for Programming the Middleware Machine
ISADS '01 Proceedings of the Fifth International Symposium on Autonomous Decentralized Systems
Dynamic Stochastic Capacity Pricing for Resource Allocation
IAT '03 Proceedings of the IEEE/WIC International Conference on Intelligent Agent Technology
Resource allocation as an evolving strategy
Evolutionary Computation
Strategies for CORBA Middleware-Based Load Balancing
IEEE Distributed Systems Online
Pricing congestible network resources
IEEE Journal on Selected Areas in Communications
Spatial pattern growth and emergent animat segregation
Web Intelligence and Agent Systems
Hi-index | 0.00 |
Following the view point of Evolutionary Dynamics, we have built a multi-agent system to study resource allocation in a heterogeneous network of resources. The class of systems we are looking at are systems facing structural uncertainties (supply structure and growth, concentration level, substitute products, ...). In our approach [1,2] resources are modeled as strategies, and agents distribute processing requirements onto resources using imperfect information and local decision making. Our agents are endowed with bounded rationality [8] and have to face the challenge of dealing with imperfect understanding of the feedback structure from resources which use unintendedly rational heuristics to set resources' unit prices. Our intent is to achieve cooperative equilibrium using competitive dynamics by controlling congestion through capacity pricing. To achieve this, a distributed differentiated pricing scheme has been used to improve loose coupling between agents and resources. The dynamics of this pricing scheme has been studied in [3]. This required a loosely coupled interaction model that adequately reflects the autonomy of the involved parties and provides the necessary spatial and temporal decoupling [4]. However, the benefits of greater decentralization and increased local decision-making come at the expense of greater stochastic dynamics which can have unpredictable effects on the stability of the system. Because such non-functional properties (stability, performance, etc) depend upon the system's underlying design and implementation [5], we had to come up with an appropriate approach for its stability analysis. This paper first describes the system under study. Following, we describe the procedure we use to analyze its stability and then show its concrete application.