A service broker for Intercloud computing

With the advantages of pay-per-use, easy access and on-demand resource customization, the Cloud computing concept was quickly adopted by both the industry and the academia. Over the last decade, the many Cloud infrastructures that have been built have distinguished themselves in the variety of offered services, access interfaces, costs and Service Level Agreement (SLA). On the other hand, “vendor lock-in” issues and the lack of common Cloud standards hinder the interoperability across Cloud providers. Thus, Cloud users have to manually make decisions about which Cloud to choose in order to meet their functional and non-functional service requirements while keeping the payment low. This task is clearly a burden for the users because they have to go through web pages of Cloud providers to compare their services and pricing policies. Furthermore, it is hard for them to collect and maintain all the needed information from current commercial Clouds to make accurate decisions. The aim of the work presented in this thesis is to facilitate users’ ability to find the most suitable Cloud services by taking their functional and non-functional SLA requirements into account. A key contribution of the thesis is the design of a multiCloud service broker framework acting as a mediator between consumers and multiple Cloud providers to automate the service selection and deployment. The framework contains components for decision-making, monitoring and SLA management, as well as an interoperability layer to interact with heterogeneous Clouds. In addition, the thesis addresses the multi-Cloud service selection problem by proposing a utility-based matching algorithm adopted from the economic theory which selects the Clouds, maximizing the user utility. To facilitate the matching of large-scale composite services, we introduce a hybrid utility-based genetic algorithm (HU-GA) that applies the user utility as an objective function. Furthermore, we present an ontological model to semantically describe the service requirements and Cloud characteristics. Moreover, the thesis investigates the optimal broker-based deployment of multi-Cloud workflows while taking into account data locality, resource heterogeneity, and task dependencies. For this purpose, a two-stage multi-dimensional resource allocation scheme is proposed, where the suitable Clouds are first selected using the HU-GA matching algorithm and then the workflow tasks are distributed to the selected Cloud