Time Shared VMs and Monitoring of Time Shared VMs

Abstract

Cloud computing is a promising technology for the implementation of scalable on-demand computing infrastructures. Self-manageable Cloud infrastructures are required in order to comply with users´ requirements specified by Service Level Agreements (SLAs) on the one hand, and to minimize user interactions with the system on the other hand. Adequate resource monitoring strategies and timely detection of possible user SLA violations represent challenging research issues. In this paper we present DeSVi-an architecture for detecting SLA violations through resource monitoring in Cloud computing infrastructures. Based on the monitored resources, the energy consumption of the application can be determined and efficiently managed. Resources are monitored using an efficient framework that is also capable of mapping lowlevel resources metrics to user-defined SLAs. The detection of possible SLA violations is based on the predefined service level objectives and we utilize knowledge databases to manage the SLA violations and the energy efficiency objectives. For the evaluation of our approach, we developed image rendering services, which exhibit heterogeneous workloads for investigating the resource (CPU, memory, storage) consumption behaviors of the applications and to determine the optimal monitoring intervals. The achieved results show that our architecture is able to monitor and detect the amount of resources consumed by the applications, which translates into the amount of energy consumed by the host.