CloudAid 2

Abstract

Since its appearance, cloud services and its market have grown remarkably. Companies have begun to consider cloud services as a real solution to their problems although, with the market growth and the appearance of an enormous quantity of cloud service providers (CSP), when one is confronted with the situation where he/she has to choose a Cloud service, discovering an ideal solution is not an easy task. Questions like “Which is the best service for my situation? or “I know what I need but there’s too many available options, which one should I choose? are hard to solve since it’s no longer a choice solely based on the price of a service. Organizations nowadays have many different types of requirements that need to be met so they can offer a product that matches the consumer’s needs. It becomes an even harder problem to solve when one is in need of a system composed of multiple services, each with different requirements. In this thesis I present an augmented version of the CloudAid1 prototype, a recommendation system that provides a suitable aggregation of cloud services given a set of requirements. This new version of the prototype aims at overcoming the limitations of the previous prototype, recurring to new Multi-Criteria Decision Methods (ELECTRE III, PROMETHEE I and SMAA-2) for higher accuracy and realism on the ordering of the decision problem’s results and, a new graph-theory based aggregation algorithm to deal with incomparability/incomplete information between alternatives. It also presents the Linked USDL Pricing API, a Java API that provides an abstraction layer between developers and the semantic models (Linked USDL core and Linked USDL pricing) thus, promoting and improving their usability. Combining the API with scrapping/parsing techniques enabled the creation of the ServiceGatherer, a simple application that supports the CloudAid2 prototype by providing semantic descriptions, of real services, for processing.