A Fuzzy-Oriented Framework For Service Ranking And Selection In Cloud E-Marketplaces

ABSTRACT

Despite the successes of commercial cloud service e-marketplaces, opportunities still

exist to improve user experience as these e-marketplaces do not yet enable dynamic

composition of atomic services to satisfy complex user requirements. More so, the

platforms employ keyword-based search mechanisms that only allow the selection of

atomic services. The elicitation mechanisms do not consider user’s QoS requirements, nor

support the elicitation of these requirements in ways akin to subjective human

expressions. In addition, search results are presented as unordered lists of icons, with

minimal comparison apparatus to simplify decision making. Existing cloud selection

approaches do not currently provide the sophistication required to optimise user

experience in the cloud e-marketplace, hence this study proposed a framework to address

the observed limitations. First, a state-of-the-art survey was conducted and six design

criteria were identified for a selection framework suitable for cloud e-marketplaces. These

criteria guided the formulation of an integrated framework, Fuzzy-Oriented Cloud

Service Selection (FOCUSS) framework. The proposed framework comprises four

modules: Cloud ecosystem and service directory, Graphical User Interface (GUI) &

Visualisation, QoS Requirement Processing, and Service Evaluation & QoS Ranking

modules. In the first module, atomic services are combined to realise the set of composite

services offered in the e-marketplace; subjective QoS requirements are then inputted via

the GUI module, and processed in the QoS requirements processing module. In service

evaluation and ranking module, the requirements are optimised and used to rank services

and the ranking results are shown to the users via bubble graph visualisation. The utility

of the proposed framework was demonstrated via a Java-based web application prototype

using a case study of a Customer Relationship Management-as-a-Service e-marketplace.

Simulation experiments and user studies were performed to evaluate the performance of

the proposed framework in terms of its scalability, ranking accuracy, and quality of user

experience. A linear regression analysis showed that the proposed framework is linearly

scalable when measured by the time it took to rank top-20 services as the number of

alternatives increased. Kruskal-Wallis and Mann-Whitney tests revealed that ranking

accuracy of proposed framework is not compromised by using subjective descriptors to

approximate user’s QoS requirements, and the ranking accuracy is higher compared to

existing approaches. Based on Wilcox signed tests, the results of the user studies showed

that users can complete tasks faster and easier compared to traditional tabular

representations. These results confirmed that the proposed framework is viable for cloud

service selection in cloud e-marketplaces. This study contributes to knowledge by

providing an integrated framework for cloud service selection that organises atomic

services within the cloud ecosystem and guides formal service composition on the fly

beyond what atomic services can deliver; handle both subjective users QoS preferences

and aspiration, and enable easy comparison of query results along multiple QoS

dimensions. In addition, it provides a framework will improve user experience, which in

turn boosts the commercial viability of cloud e-marketplaces.