Energy-efficient and SLA-Aware Management of IaaS Clouds

Cloud computing utilizes arbitrary mega-scale computing infrastructures and is currently revolutionizing the ICT landscape by allowing remote access to computing power and data over the Internet. Besides the huge economical impact Cloud technology exhibits a high potential to be a cornerstone of a new generation of sustainable and energy-efficient ICT. The challenging issue thereby is the energy-efficient utilization of physical machines (PMs) and the resource-efficient management of virtual machines (VMs) while attaining promised non-functional qualities of service expressed by means of Service Level Agreements (SLAs). Currently, there exist solutions for PM power management, VM migrations, and dynamic reconfiguration of VMs. However, most of the existing approaches consider each of them alone, and only use rudimentary concepts for migration costs or disrespect the nature of the highly volatile workloads. In this paper we present an integrated approach for VM migration and reconfiguration, and PM power management. Thereby, we incorporate an autonomic management loop, where proactive actions are suggested for all three areas in a hierarchically structured way. We evaluate our approach with both, synthetic workload data and real-word monitoring data of a Next Generation Sequencing (NGS) application used for the protein folding in the bioinformatics area. The efficacy of our approach is evaluated by considering classical algorithms like First Fit, Monte Carlo and Vector Packing, adapted for energy-efficient reallocation. The results show energy savings up to 61.6% while keeping acceptably low SLA violation rates.