Three-Tier Grid Architecture with Policy-Based Resource Selection in Grid

Abstract

Trust management for the selection of resources is a key issue in a grid environment. In most of the trust-based systems, intermediaries, known as brokers, are responsible for the selection of Service Providers (SPs) for consumer requests. In such models, the intermediaries gain monetary benefit for each of the transactions made through them. This may lead to favoritism & bias in the selection of suitable SPs. This challenge may be overcome by the proposed three-tier framework where the Regional Resource Administrators (RRAs), resource brokers and the SPs are arranged into three-tiers. An RRA selects a 'suitable' SP for a consumer request, based on the trust-indices and Quality of Services of the SPs nominated by various brokers. RRAs derive their compensation from registration, renewal, and audit charges paid by the broker community, not from individual transactions. RRAs serve the consumer community in a manner similar to DNS, in an 'unbiased' and 'trustworthy' manner. In this paper, our proposed architecture not only supports a choice of SP based on reputation (trust-index) but also on credentials (policy). Thereby the consumer is assigned with a trustworthy SP and the transaction is free from runtime failure as the policies have been matched with. The concept of de-registration has been introduced in two scenarios, the SPs may deregister themselves by choice and the other is the RRAs force brokers to deregister. In this model, the trust-indices of each of the entities (consumers, SPs and brokers) arc computed based on the feedback provided by other entities after each transaction. These trust-indices of brokers, consumers and SPs are updated dynamically at the RRA's and the broker's sites respectively, to ensure trustworthy services and to quicken the selection of 'suitable' SPs. Our model shows a marked improvement in job-success-rate for various percentages of malicious entities.