English

Abstract

Cloud computing is a promising concept for the implementation of scalable on-demand computing infrastructures, where resources are provided in a self-managing manner based on predefined customers requirements. A Service Level Agreement (SLA), which is established between a Cloud provider and a customer, specifies these requirements. It includes terms like required memory consumption, bandwidth or service availability. The SLA also defines penalties for SLA violations when the Cloud provider fails to provide the agreed amount of resources or quality of service. A current challenge in Cloud environments is to detect any possible SLA violation and to timely react upon it to avoid paying penalties, as well as reduce unnecessary resource consumption by managing resources more efficiently. In resource-shared Cloud environments, where there might be multiple VMs on a single physical machine and multiple applications on a single VM, Cloud providers require mechanisms for monitoring resource and QoS metrics for each customer application separately. Currently, there is a lack of generic classification of application level metrics. In this paper, we introduce a novel approach for classifying and monitoring application level metrics in a resource-shared Cloud environment. We present the design and implementation of the generic application level monitoring system. Finally, we evaluate our approach and implementation, and provide a proof of concept and functionality.