A Framework for managing Quality of Service in Cloud Computing through Service Level Agreements

Until today, Cloud providers only offer a limited set of non-negotiable service levels to their customers. Most often these service levels are expressed as guarantees for availability together with the offer to have access to a virtualised environment with a certain performance the customer may select from a number of predefined configurations. This simplifies the life of the provider, e.g., in terms of effort to maintain an adequate infrastructure, or regarding the effort for reducing the risk violating Service Level Agreements (SLAs) with its customer. In consequence, the current practice is slanted towards the benefit of the provider and ignores more specific requirements of its customers, e.g. regarding data protection and related guarantees. An analysis of the underlying problems shows two major fields to be worked on for solving the problem: Firstly, each provider uses its own proprietary technology for managing SLAs throughout their life-cycles. However, first standards are available and could be employed allowing the customer to use a single standard interface to negotiate with several providers. Secondly, there is neither a common set of terms to describe Cloud customers requirements regarding the Cloud services requested, nor, the back side of the medal, there is a common set of terms to describe the Quality of Service (QoS) of the Cloud providers' offerings. The focus of the presented work is (i) on the standard technology for negotiating and creating SLAs and (ii) the common terms and metrics describing providers' offerings and customers' requirements. Without these terms mapping the customers' requirements to cloud providers' offerings is a tedious manual and error-prone process and resulting SLAs will remain rudimentary. Additionally, both providers and their customers would benefit from more sophisticated and negotiable Service Level Agreements using existing standards. These SLAs are both (i) binding and monitorable agreements between the customer and the provider covering the customers' requirements and (ii) the basis for a QoS-aware Cloud resource management on the side of the provider including provisioning of physical machines and optimised allocation of virtual machines. Besides more traditional QoS aspects, terms related to Cloud Federation, Data Protection or Security Level Agreements are covered. Customers may use standards compliant agreement templates with the providers' offerings to select suitable providers for starting the negotiation to the extent the provider allows in the template. During the negotiation the provider may take into account the actual degree of capacity utilisation of its infrastructure and active SLAs, and may use the SLA resulting from a successful negotiation to further optimise its infrastructure through application and VM consolidation. The work presented in this thesis covers about a decade of research starting in the environment of Grid computing and ending with today's Cloud computing.