Green vs revenue: Data center profit maximization under green degree constraints

Huaiwen He, Hong Shen

研究成果: Conference contribution同行評審

摘要

With the soaring personal and enterprise computation demands, the scale of cloud centers has been rapidly increasing, which leads to massive amounts of greenhouse gas emission. From the cloud service provider’s (CSP) perspective, profit is the key factor to maintain the development of cloud centers, which potentially conflicts with the goal of achieving green data centers for environment protection because the expensive renewable energy will add more cost. This paper addresses the problem of maximizing profit while meeting the green degree constraints for a large scale data center on renewable energy source. Taking into account of the bursty randomness of workload, time-varying electricity price and intermittent green energy, we first formulate the problem in an optimization framework with stochastic constraints for delay-tolerant workload. Then, we show how the deployment of Lyapunov optimization technique can leverage to obtain a low-complexity online solution for profit maximization by combining request admission control, workload scheduling and power management. Moreover, we adopt a non-linear submodular revenue function to optimize the throughput of the system. By decoupling the optimization function of a time average problem into three sub-problems, we solve them to obtain the optimal control strategies. Our proposed algorithm achieves a desirable profit-green tradeoff. At the end, we provide the performance bound of our algorithm, and evaluate its performance through extensive trace-driven simulations.

原文English
主出版物標題Parallel and Distributed Computing, Applications and Technologies - 19th International Conference, PDCAT 2018, Revised Selected Papers
編輯Hui Tian, Jong Hyuk Park, Yunsick Sung, Hong Shen
發行者Springer Verlag
頁面28-40
頁數13
ISBN(列印)9789811359064
DOIs
出版狀態Published - 2019
對外發佈
事件19th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2018 - Jeju Island, Korea, Republic of
持續時間: 20 8月 201822 8月 2018

出版系列

名字Communications in Computer and Information Science
931
ISSN(列印)1865-0929

Conference

Conference19th International Conference on Parallel and Distributed Computing, Applications and Technologies, PDCAT 2018
國家/地區Korea, Republic of
城市Jeju Island
期間20/08/1822/08/18

指紋

深入研究「Green vs revenue: Data center profit maximization under green degree constraints」主題。共同形成了獨特的指紋。

引用此