A Self-regulating System for Resource Sharing in Multi-site Grid Environments

We describe a self-adaptive and self-regulating system for fair sharing of resources in a multi-site grid environment. In such an environment, each site has a set of resources and has internal demand which it manages using internal policies and rules. To manage periodic fluctuations in the local demand and potential shortfall of resources to meet the local demand, the participants borrow and lend resources to one another. This resource sharing among the participants is regulated by a distributed, peer-to-peer resource brokering algorithmthat is scalable and fair in its allocation of the shared resources. The self-regulatingmechanism enables discovery of shared resources and the possible extent of the donated or borrowed resources. The regulating mechanism automatically clusters participants with similar borrowing and donating properties. We present both analytical and empirical evidence to prove the properties of such a system. In particular, we prove that the system is fair – that is, participants who contribute more resources to the shared pool get to borrow proportionately more resources from others. Such a resource brokering mechanism allows participants in a multi-site grid environment to meet their peak demands by provisioning resources only for average demand.