Using Multi-Agent Principles for Implementing an Organic Real-Time Middleware

Abstract

This article describes the architecture of an organic, real-time capable, service-based middleware. We show how middleware services can be treated as intelligent agents, and how we can use a multi-agent coordination mechanism for implementing the organic management, in particular self-configuration and self-optimization, in a decentralized and efficient way.

Guiding Organic Management in a Service-Oriented Real-Time Middleware Architecture

“…It is possible to specify both dependencies between services and dependencies on resources. By combining the capability mechanism with an auction-based task allocation mechanism as described in [7], self-optimization, self-configuration and self-healing can be achieved. We are currently implementing the proposed mechanism in our middleware CARISMA, which is part of the CAR-SoC project [13].…”

“…The general architecture for an organic, service-oriented middleware inspired by the OSA+ approach has been developed in [12]. For this architecture, we described an agent-based approach for implementing self-X properties in [7], which uses concepts from multi-agent systems for coordination and task allocation. A short summary of our approach is given in Sect.…”

“…In such an architecture, tasks are processed by services, which in most cases are not part of the middleware core, but run as independent, loosely coupled entities, leading to a microkernel design. We have proposed and explained in detail such a design in [7]; here, we will only give a short overview. We have chosen a microkernel-based approach primarily for scalability, flexibility, reliability, recovery and extensibility.…”

“…In [7], we have proposed a service-oriented organic real-time middleware architecture that achieves self-X properties through a multi-agent-based approach. Services act as intelligent agents that use an auction mechanism to coordinate.…”