Planning and Optimization During the Life-Cycle of Service Level Agreements for Cloud Computing

Abstract

A Service Level Agreement (SLA) is an electronic contract between the consumer and the provider of a service. It governs their business relationship by clarifying expectations and obligations of participating entities, with regard to the service and its quality. SLAs are already the prime paradigm for the description of cloud computing services. Once an SLA is established, the provider has to ensure that service quality remains within certain acceptable levels; and comply with the customer's demands until the end of the service life time. However, managing the SLAs is still a technical challenge that requires signi cant e ort to achieve autonomy, economy and e ciency. Current state-of-the-art in SLA management faces challenges such as SLA representation for cloud services; business-related SLA optimizations; service outsourcing and resource management. These areas constitute, as one would expect, major contemporary research topics. Hence, a structured methodology engineered for the management of the di erent phases of SLAs during its lifespan is of paramount importance, which indeed facilitates the realization of the cloud SLA management. To this aim, I present diversi ed models and approaches in SLA lifecycle management that address the aforementioned challenges and enable automatic service modeling, negotiation, provisioning and monitoring. During the SLA creation phase, I outline how to improve and simplify the structures that model SLAs. Furthermore, another objective of my approach is to minimize implementation and outsourcing costs for reasons of competitiveness, while respecting business policies for pro t and risk. During the SLA monitoring phase, I develop the strategies for virtual cloud resources selection and allocation during live migrations. Then, I apply an appropriate theoretical model for ne-grained yet simpli ed and practical monitoring of massive sets of SLAs, that separates the agreement's fault-tolerance concerns into multiple autonomous layers. The work at hand contributes a blueprint for the GWDG and its scienti c communities. The research that lead to this thesis was conducted as part of the SLA@SOI EU/FP7 Integrated Project (contract No. 216556).