Managing Cloud Service Provisioning and SLA Enforcement via Holistic Monitoring Techniques

Rapid technological developments in Information Technology (IT) and ubiquitous Internet access are causing serious challenges in service provisioning and resource management landscapes. Cloud computing is proving to be a reliable technology to address these challenges. Service provisioning in the Cloud relies on Service Level Agreements (SLAs) representing a contract signed between the customer and the service provider including non-functional requirements of the service specified as Quality of Service (QoS) and penalties in case of violations. Flexible and reliable management of resources and SLA agreements are of paramount importance to both Cloud providers and consumers. On the one hand, providers have to prevent SLA violations to avoid penalties and on the other hand, they have to ensure high resource utilization to prevent costly maintenance of unused resources. Although, there is a large body of work considering development of flexible and self-manageable Cloud computing infrastructures, there is still a lack of adequate monitoring infrastructures capable of predicting possible SLA violations. Most of the available monitoring systems rely either on Grid or service-oriented infrastructures, which are not directly compatible to Clouds due to the differences in resource usage models, or due to heavily network-oriented monitoring infrastructures. Furthermore, to achieve high resource utilization and more revenue, the providers must be able to schedule resources and deploy different user applications complying with the SLA objectives, and at the same time optimizing the performance of the applications. The current scheduling approaches in Clouds are tailored toward the usage of single SLA objectives, such as execution time in decision making. The design of a generalized scheduling algorithm for optimal mapping of an application with multiple SLA parameters to resources is yet to be investigated. Nevertheless, the idea of scheduling and deploying multiple applications on the same host brings a new set of challenges to the Cloud provider because he must enforce the SLA terms of each customer application independently. In this thesis, we propose a novel Cloud management infrastructure, which is based on holistic monitoring techniques and mechanisms for low-level resource metrics to high-level SLA mapping, application scheduling and deployment, and the ability to monitor multiple application executing on the same host. We present the design and implementation of these techniques. In a case study, we show the integration of knowledge management techniques into Cloud management infrastructures realizing autonomic behaviour and providing reactive action to prevent / correct the SLA violation situations. Finally, we present some evaluations to show the novelty of the contributed solutions.