Decentralized Task Allocation in an Organic Real-Time Middleware - An Auction-Based Approach

In the domain of distributed, embedded, real-time systems, task allocation is an important problem to consider. In autonomic/organic computing (AC/OC), systems shall autonomously configure and constantly optimize themselves. In this context, finding an efficient and decentralized task allocation mechanism is essential. We are currently developing an autonomic middleware within the CAR-SoC project. It has a service-oriented architecture, where services are agents which behave independently and intelligently. Using a mechanism well known from the multi-agent systems realm, they autonomously allocate tasks among themselves, not requiring a central mediator. In this paper, we present our approach, the current state of our implementation, and an example where we apply our mechanism to a load-balancing scenario.