Modelling the performance of web services in cloud e-marketplaces based on consumer waiting time and provider cost

Abstract

Cloud E-Marketplaces are virtualised global network markets that allow the exchange of digital information through a broker for the purpose of conducting and delivering cost effective business services. Prior to the Cloud, E-Marketplaces were the Traditional Internet web service and Grid E-Markets. All these Marketplaces experienced challenges which led to the creation of Cloud E-Marketplaces for service delivery in form of Software, Platform and Infrastructure. Research on Cloud E-Marketplaces has concentrated on sub-domains including security [1], energy [2] and privacy [3] but little has been done with regard to optimization for resource management in Cloud performance [4]. One major parameter that is required to bring the cost down is the number of server machine(s) used by a service provider during service provisioning. A second one is the performance measures like waiting time which are used to determine the effectiveness of Cloud E-Marketplaces’ performance. In order to establish the thesis that minimizsation of server machine cost and consumer waiting time in the context of non-priority and non-pre-emptive priority policy is imperative, the study accomplished the following: i. It extensively reviewed the existing body of knowledge on the performance of E-Marketplaces. ii. It identified the need to re-engineer the existing Cloud E-Marketplace architecture as networks of queues with parallel web stations with a feedback from scheduler in the context of non-priority and non-pre-emptive policy without dedicating any web station to any class to achieve optimal service level. iii. It evaluated the Non Priority First Come First Serve, FCFS service discipline and the Non-Preemptive model in order to gauge the performance impact on consumers’ waiting time and providers’ cost. iv. It formulated a cost structure that balances the server machine (Service Level) and consumers’ waiting time on both non-priority and non-preemptive models. v. It formulated a dynamic waiting time optimisation control mechanism that further addressed the issues of service over and under-provisioning.

The contributions are: i. the evaluative study of non-priority queues in series against the generalised approach that uses a single point of entry as proposed by others in the literature. This was used to determine the optimal service level and consumer waiting time. ii. the exhaustive evaluation of a novel non-preemptive architectural model of the Cloud E-Marketplace with each of service stations modeled as M/M/c/Pr against the M/M/1 proposed in the literature. This model was unique in that it: i. explored a different mathematical and simulation concept and also; ii. resolved the challenge of dedicating or allocating servers to a particular consumer class thereby reducing consumers’ waiting time. iii. investigated E-Marketplaces under the non-priority and also the two service non pre-emptive and the generalised models; iv. introduced the novel concept of profitability and Cost Benefit Ratio by using the Dynamic Control Model (DCM) over the Fixed Server Model (FSM).