A dynamic energy conservation scheme with dual-rate adjustment and semi-sleep mode in cloud system
Yu Cui,Yang Zhang,Xia Li,Shunfu Jin
DOI: https://doi.org/10.1007/s11227-022-04715-w
IF: 3.3
2022-08-16
The Journal of Supercomputing
Abstract:In view of the constantly rising energy cost and increasingly stringent environmental standards, improving energy efficiency and reducing carbon emission are the key to the sustainable development of cloud computing. Based on dual-rate adjustment and semi-sleep mode, in this paper, we propose a dynamic energy conservation scheme in cloud system. Firstly, we construct a cloud system with two types of physical machines (PMs) called the hot (continuous running) PM and warm (turned on, but in a dynamic sleep) PM, respectively. Each PM is deployed with multiple virtual machines (VMs) and a resource search engine (RSE). In the hot PM, a dual-rate adjustment mechanism (operating the running rates of the VMs and RSE between high and low speeds) is introduced. In the warm PM, a semi-sleep mode (switching the VMs and RSE between normal and semi-sleep states) is employed, where semi-sleep means running at lower rate rather than stopping working. For the proposed energy conservation scheme, we build a hybrid queueing model with adaptive service rate and synchronous multi-working-vacation. Using the quasi-birth-and-death (QBD) process and matrix-geometric solution, we derive the average waiting time of requests and energy saving rate of system. Through numerical results, we reveal the influence of dual-rate adjustment and semi-sleep mode on the system performance, and verify the effectiveness of our proposed scheme in improving system performance. Finally, from the perspective of economics, we establish a cost function of system to compromise different performance measures. With the goal of minimizing the system cost, we develop an improved Salp Swarm Algorithm (SSA) to optimize the system performance.
computer science, theory & methods,engineering, electrical & electronic, hardware & architecture