A time-constrained SLA negotiation strategy in competitive computational grids

Automated and intelligent negotiation solutions for reaching service level agreements (SLA) represent a hot research topic in computational grids. Previous work regarding SLA negotiation in grids focuses on devising bargaining models where service providers and consumers can meet and exchange SLA offers and counteroffers. Recent developments in agent research introduce strategies based on opponent learning for contract negotiation. In this paper we design a generic framework for strategical negotiation of service level values under time constraints and exemplify the usage of our framework by extending the Bayesian learning agent to cope with the limited duration of a negotiation session. We prove that opponent learning strategies are worth for consideration in open competitive computational grids, leading towards an optimal allocation of resources and fair satisfaction of participants.

► We study SLA negotiation strategies in computational grids with open participation.
► We propose a framework to build time-constrained strategies starting from opponent learning baseline strategies.
► We adapt the Bayesian learning negotiation strategy to cope with time constraints.
► We show that the obtained outcome assures optimal allocation of resources and fair satisfaction of participants.