Abstract
Cloud data centers (CDCs) with abundant resource capacities have prevailed in the past decade. However, these CDCs often struggle to efficiently deal with resource provisioning in terms of performance and energy efficiency. In this paper, we present Energy-Based Auto Scaling (EBAS) as a new resource auto-scaling approach-that takes into account Service Level Agreement (SLA)-for CDCs. EBAS proactively scales resources at the CPU core level in terms of both the number and frequency of cores. It incorporates the dynamic voltage and frequency scaling (DVFS) technique to dynamically adjust CPU frequencies. The proactive decisions on resource scaling are enabled primarily by the CPU usage prediction model and the workload consolidation model of EBAS. The experimental results show that EBAS can save energy on average by 14% compared with the Linux governor. In particular, EBAS contributes to enhancing DVFS by making it aware of SLA conditions, which leads to savings of computing power and in turn energy.