Along with manifold advantages of distributed multi-agent systems, increased network traffic produced by highly communicative agents in large distributed systems has to be considered as their practical downside. We suggest to apply the programming paradigm of mobile agents to reduce this network overhead by allowing agents to meet at the same network node before commencing communication. A remote communication between two agents could then be replaced by one or two agent migrations, followed by local communication. Since only in trivial cases it is possible to decide at design time whether remote communication or agent migration with subsequent local communication would perform better, this decision has to be made at run-time based on environmental parameters and agents' past experience. We present an adaptive approach, which is inspired by a solution of the El Farol problem. Every agent forecasts the network load of the next communication step and applies a simple mathematical model to decide between the two alternatives at run-time. In addition, our approach does not only consider network load but also server load by enabling agents to dynamically forecast the number of agents migrating to a specific agent server. The approach is evaluated with simulation experiments in static and dynamic server load environments.
This series reports new developments in agent-based software technologies and agentoriented software engineering methodologies, with particular emphasis on applications in various scientific and industrial areas. It includes research level monographs, polished notes arising from research and industrial projects, outstanding PhD theses, and proceedings of focused meetings and conferences. The series aims at promoting advanced research as well as at facilitating know-how transfer to industrial use.
About Whitestein TechnologiesWhitestein Technologies AG was founded in 1999 with the mission to become a leading provider of advanced software agent technologies, products, solutions, and services for various applications and industries. Whitestein Technologies strongly believes that software agent technologies, in combination with other leading-edge technologies like web services and mobile wireless computing, will enable attractive opportunities for the design and the implementation of a new generation of distributed information systems and network infrastructures.
The vision of transparent, on-demand resource utilization in distributed and open environments requires resource management techniques that are robust, scalable and able to adapt to the dynamic environment. In this paper, we propose a decentralized resource allocation algorithm for the co-allocation of interrelated resources for repeated jobs in real-time. Resource broker agents autonomously allocate resources for the execution of jobs based on information from past allocations. The coordination between agents for an overall efficient resource allocation emerges through individual feedback that agents receive from the quality of previous resource allocation decisions. The coordination between agents is achieved without any communication. We present experimental results demonstrating that the proposed algorithm is able to adapt to the dynamic environment for an efficient utilisation of the system resources.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.