Using Information Filtering in Web Data Mining Process

Abstract

Web service-oriented Grid is becoming a standard for achieving loosely coupled distributed computing. Grid services could easily be specified with web-service based interfaces. In this paper we first envisage a realistic Grid market with players such as end-users, brokers and service providers participating co-operatively with an aim to meet requirements and earn profit. End-users wish to use functionality of Grid services by paying the minimum possible price or price confined within a specified budget, brokers aim to maximise profit whilst establishing a SLA (Service Level Agreement) and satisfying end-user needs and at the same time resisting the volatility of service execution time and availability. Service providers aim to develop price models based on end-user or broker demands that will maximise their profit. In this paper we focus on developing stochastic approaches to end-user workflow scheduling that provides QoS guarantees by establishing a SLA. We also develop a novel 2-stage stochastic programming technique that aims at establishing a SLA with end-users regarding satisfying their workflow QoS requirements. We develop a scheduling (workload allocation) technique based on linear programming that embeds the negotiated workflow QoS into the program and model Grid services as generalised queues. This technique is shown to outperform existing scheduling techniques that don't rely on real-time performance information.