A proposed energy and performance aware cloud framework for improving service level agreements (SLAs) in cloud datacenters

Physical computer hardware is being replaced with virtual hardware in cloud computing for cost efficient operations. It is expected that by 2020, most of medium and large organizations will migrate to cloud computing for enhanced and sustainable business. However, the trade-off between meeting Service Level Agreements (SLA) and minimizing energy consumption of physical machines in data centers is not fully optimized, thus leaving scope for further improvement. In spite of the fact that cloud providers are using state-of-the-art technologies, cloud clients are still demanding higher and ever increasing Quality of Service (QoS) to satisfy the need of customers. Based on the authors’ previous practical work on a high-end server on ESXi 5.5 hypervisor platform, a novel framework for improving computational efficiency, performance and reducing cloud energy consumption is proposed in this paper. The proposed framework integrates with hardware through a server classification process (idle server, under-loaded server, balance server and over-loaded server) and distributes computational loads with a built-in logic to reduce energy consumption. In addition, the framework promises an efficient solution for Virtual Machines (VMs) allocation and optimization that will satisfy SLAs for cloud consumers. In all four server categories, tracking and a recording system is considered for Physical Machine (PMs) and VMs. For effective utilization of the idle server state, a wake-up and sleep mode decider are proposed. The uniqueness of the framework can be validated with its implementation on CloudSim softwar