Beyond user ranking: Expanding the definition of reputation in grid computing

Shopping around for a good service provider in a Grid Computing environment is no less challenging than the traditional shopping around in non-virtual marketplace. A client may consult a service broker for providers that can meet specific QoS requirements (e.g., CPU speed), and the broker may return a list of candidate providers that satisfy the client's demands. If this computing platform is backed up by some reputation system, the list of providers is then sorted based on some reputation criterion, which is commonly the user rating. We argue in this paper that judging the reputation of a provider based on user rating is not sufficient. The reputation should additionally reflect how trustworthy that provider has been with respect to complying with the finalized SLA (using a metric called conformance) and how consistent it has been with respect to honouring its compliance levels (using a metric called fidelity). Accordingly, we perceive the reputation as a vector of three dimensions: user rating, conformance, and fidelity. In this paper, we define these metrics, explain how to compute them formally, and how to use them in the reputation-enabled framework that we describe.