Negotiation process for resource allocation in grid using a multi-agent system

Grid technology is used for large-scale resource sharing among different institutions or individuals. A recent challenge in this environment is to provide guarantees for the service delivery in a job submission. It is important to know when a job will finish and the performance and costs involved. We present here a multi-agent system that considers such aspects to decide about the best place to run a job in grid. The main component of our solution is the negotiation process, which relies on resource access policies and performance prediction based on past cases of job submissions. The negotiation final goal is to establish Service Level Agreements that register user's expectations and service provision according to price, quality, performance and scheduling.