Dynamic allocation of power delivery paths in consolidated data centers based on adaptive UPS switching

被引:23
作者
Al-Hazemi, Fawaz [1 ]
Peng, Yuyang [1 ]
Youn, Chan-Hyun [1 ]
Lorincz, Josip [2 ]
Li, Chao [3 ]
Song, Guo [4 ]
Boutaba, Raouf [5 ]
机构
[1] Korea Adv Inst Sci & Technol, Sch Elect Engn, Daejeon, South Korea
[2] Univ Split, Fac Elect Engn, Mech Engn & Naval Architecture FESB, Split, Croatia
[3] Shanghai Jiao Tong Univ, Dept Comp Sci & Engn, Shanghai, Peoples R China
[4] Hong Kong Polytech Univ, Dept Comp, Hong Kong, Hong Kong, Peoples R China
[5] Univ Waterloo, Fac Comp Sci, Waterloo, ON, Canada
关键词
Data center; Energy efficiency; Power management; UPS; VIRTUAL MACHINES; ENERGY;
D O I
10.1016/j.comnet.2018.08.004
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Although technique known as server consolidation approach in a data center can reduce the overall power consumption, the Power Usage Effectiveness (PUE) of the data center will still be negatively affected with presence of distributed Uninterruptible Power Supplies (UPSs). The impact on the PUE arises from the fact that all UPS modules are kept running to maintain power availability for only a few active servers during off-peak periods. To address this problem, in this paper technique for reducing power consumption in a data center by consolidating the UPSs used during off peak periods is proposed. The proposed technique achieves power savings by leveraging a micro Automatic Transfer Switch (micro-ATS) at the server end. The novelty of this work lies in developed adaptive algorithm that continuously looks for opportunities to reduce the number of UPSs by offloading under-loaded UPSs to a neighboring UPS whenever that neighboring UPS can handle the extra load. In various simulated scenarios involving corporate data centers, our approach demonstrates the ability to save more power and achieve lower PUE degradation compared with state-of-the-art approaches such as server consolidation. Specifically, the proposed approach achieves a savings of approximately 20% to 40% in a data center's power consumption, depending on the data center's off-peak periods, which can be accomplished using only 80% of the UPS modules in the data center. (C) 2018 Elsevier B.V. All rights reserved.
引用
收藏
页码:254 / 270
页数:17
相关论文
共 62 条
[1]  
Aksanli B., 2013, GREEN COMP C IGCC 20, P1
[2]   Architecting Efficient Peak Power Shaving Using Batteries in Data Centers [J].
Aksanli, Baris ;
Rosing, Tajana ;
Pettis, Eddie .
2013 IEEE 21ST INTERNATIONAL SYMPOSIUM ON MODELING, ANALYSIS & SIMULATION OF COMPUTER AND TELECOMMUNICATION SYSTEMS (MASCOTS 2013), 2013, :242-+
[3]   A MISO model for power consumption in virtualized servers [J].
Al-Hazemi, Fawaz ;
Peng, Yuyang ;
Youn, Chan-Hyun .
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS, 2015, 18 (02) :847-863
[4]  
Alharbi F., 2016, Complex and Intelligent Systems, V2, P155, DOI [10.1007/s40747-016-0021-9, DOI 10.1007/S40747-016-0021-9]
[5]  
[Anonymous], 2014, MSRTR201437
[6]  
[Anonymous], [No title captured], DOI DOI 10.1093/BIB/BBW042
[7]  
[Anonymous], 2010, DAT CTR SIT INFR TIE
[8]  
[Anonymous], 2011, The Matlab Users Guide
[9]  
[Anonymous], 2014, UPTIME I REPORT
[10]  
[Anonymous], 2009, ARXIV09032525