Configuration management for multi-agent systems

Abstract

As heterogeneous distributed systems, multi-agent s y s t e m s present some challenging con guration management issues.There are the problems of knowing how to allocate agents to computers, launch them on remote hosts, and once the agents have been launched, how to monitor their runtime status so as to manage computing resources e ectively.In this paper, we present the RETSINA Con guration Manager, RECoMa. W e describe its architecture, how i t uses agent infrastructure such as service discovery, to assist the multi-agent system administrator in allocating, launching, and monitoring a heterogeneous distributed agent s y stem in a distributed and networked computing environment. 1

A Genetic Approach for Long Term Virtual Organization Distribution

“…33 In Giampapa, et al, a graphical interface is presented which helps human administrators to launch agents based on their software requirements and the requirements offered by the different platform hosts. 19 In that sense, the idea is similar to the way we propose to handle agent requirements. However, these operations are launched by a human operator and they do not seek to automatically balance workload on hosts.…”

“…32 Several MAS platforms include management tools that allow agents to be distributed across platforms hosts. 19,32,33,34,35 Sanchez-Anguix, et al, 32 propose a multi-agent management architecture based on the Magentix platform. 36 It provides a graphical tool which allows several management operations such as MAS visualization, agent creation, agent removal, and MAS' history visualization.…”

“…Additionally, open MAS make also reasonable to assume that different agents may have different software requirements (operating system, software libraries, etc.). 19 Consequently, these requirements are to be taken into account when distributing agents across shared resources.…”

A Multi-agent Framework for Dependable Adaptation of Evolving System Architectures

“…An agent handles reconfiguration requests for its associated component. RECoMa [27] is a configuration manager can help find appropriate computer platforms to deploy software agents of a multi-agent system. All available devices on which available components can be deployed are treated as equal, each capable of deploying the components allocated to it.…”

Approximation results for probabilistic survivability

“…They use teamwork theory to specify how to recover from broker failure. The RecoMa system [7] uses multiple servers to support matching agents to computers. None of these works assume a probabilistic failure model for nodes and they do not optimize a probabilistic objective function that captures survivability.…”