融合XGBoost和Multi-GRU的数据中心服务器能耗优化算法

doi:10.11772/j.issn.1001-9081.2021071291

《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (1): 198-208.DOI: 10.11772/j.issn.1001-9081.2021071291

• 先进计算 • 上一篇

融合XGBoost和Multi-GRU的数据中心服务器能耗优化算法

申明尧, 韩萌(), 杜诗语, 孙蕊, 张春砚

北方民族大学计算机科学与工程学院，银川 750021

收稿日期:2021-07-19 修回日期:2021-09-03 接受日期:2021-09-15 发布日期:2021-09-03 出版日期:2022-01-10
通讯作者: 韩萌
作者简介:申明尧（1994—），男，山东菏泽人，硕士研究生，CCF会员，主要研究方向：数据挖掘、人工智能
韩萌（1982—），女，河南商丘人，副教授，博士，CCF会员，主要研究方向：数据挖掘
杜诗语（1996—），女，辽宁抚顺人，硕士，主要研究方向：数据挖掘
孙蕊（1993—），女，山东济宁人，硕士，主要研究方向：数据挖掘
张春砚（1995—），女，河北张家口人，硕士研究生，主要研究方向：数据挖掘。
基金资助:
国家自然科学基金资助项目(62062004);宁夏自然科学基金资助项目(2020AAC03216)

Data center server energy consumption optimization algorithm combining XGBoost and Multi-GRU

Mingyao SHEN, Meng HAN(), Shiyu DU, Rui SUN, Chunyan ZHANG

School of Computer Science and Engineering，North Minzu University，Yinchuan Ningxia 750021，China

Received:2021-07-19 Revised:2021-09-03 Accepted:2021-09-15 Online:2021-09-03 Published:2022-01-10
Contact: Meng HAN
About author:SHEN Mingyao， born in 1994， M. S. candidate. His research interests include data mining， artificial intelligence.
HAN Meng， born in 1982， Ph. D.， associate professor. Her research interests include data mining.
DU Shiyu， born in 1996， M. S. Her research interests include data mining.
SUN Rui， born in 1993， M. S. Her research interests include data mining.
ZHANG Chunyan， born in 1995， M. S. candidate. Her research interests include data mining.
Supported by:
National Natural Science Foundation of China(62062004);Natural Science Foundation of Ningxia(2020AAC03216)

摘要/Abstract

摘要：

随着云计算技术的快速发展，数据中心的数量大幅增加，随之而来的能源消耗问题逐渐成为一个研究热点。针对服务器能耗优化问题，提出了一种融合极限梯度提升（XGBoost）和多个门控循环单元（Multi-GRU）的数据中心服务器能耗优化（ECOXG）算法。首先利用Linux终端监控命令和功耗仪收集服务器各部件的资源占用信息和能耗等数据，并对其进行数据预处理来得到资源利用率。其次将资源利用率串联构造成向量形式的时间序列，用其训练Multi-GRU负载预测模型，并根据预测结果对服务器进行模拟降频，以得到降频后的负载数据。然后将服务器的资源利用率与相同时刻的能耗数据相结合，并用其训练XGBoost能耗预测模型。最后将降频后的负载数据输入到训练后的XGBoost模型中，从而预测出降频后的服务器能耗。在6台物理服务器实际资源利用率数据上的实验表明，与卷积神经网络（CNN）、长短期记忆（LSTM）网络、CNN-GRU和CNN-LSTM模型相比，ECOXG算法在均方根误差（RMSE）上分别降低了50.9%、31.0%、32.7%、22.9%；同时，与LSTM、CNN-GRU和CNN-LSTM模型相比，ECOXG算法在训练时间上分别节省了43.2%、47.1%、59.9%。实验结果表明，ECOXG算法能够在服务器能耗预测和能耗优化方面提供一定的理论基础，且在准确性和运行效率方面明显优于对比算法。此外，模拟降频后的服务器能耗已明显低于真实能耗，且在服务器的利用率较低时降耗效果显著。

关键词: 数据中心, 能耗优化, 负载, 极限梯度提升, 多个门控循环单元

Abstract:

With the rapid development of cloud computing technology， the number of data centers have increased significantly， and the subsequent energy consumption problem gradually become one of the research hotspots. Aiming at the problem of server energy consumption optimization， a data center server energy consumption optimization combining eXtreme Gradient Boosting （XGBoost） and Multi-Gated Recurrent Unit （Multi-GRU）（ECOXG） algorithm was proposed. Firstly， the data such as resource occupation information and energy consumption of each component of the servers were collected by the Linux terminal monitoring commands and power consumption meters， and the data were preprocessed to obtain the resource utilization rates. Secondly， the resource utilization rates were constructed in series into a time series in vector form， which was used to train the Multi-GRU load prediction model， and the simulated frequency reduction was performed to the servers according to the prediction results to obtain the load data after frequency reduction. Thirdly， the resource utilization rates of the servers were combined with the energy consumption data at the same time to train the XGBoost energy consumption prediction model. Finally， the load data after frequency reduction were input into the trained XGBoost model， and the energy consumption of the servers after frequency reduction was predicted. Experiments on the actual resource utilization data of 6 physical servers showed that ECOXG algorithm had a Root Mean Square Error （RMSE） reduced by 50.9%， 31.0%， 32.7%， 22.9% compared with Convolutional Neural Network （CNN）， Long Short-Term Memory （LSTM） network， CNN-GRU and CNN-LSTM models， respectively. Meanwhile， compared with LSTM， CNN-GRU and CNN-LSTM models， ECOXG algorithm saved 43.2%， 47.1%， 59.9% training time， respectively. Experimental results show that ECOXG algorithm can provide a theoretical basis for the prediction and optimization of server energy consumption optimization， and it is significantly better than the comparison algorithms in accuracy and operating efficiency. In addition， the power consumption of the server after the simulated frequency reduction is significantly lower than the real power consumption， and the effect of reducing energy consumption is outstanding when the utilization rates of the servers are low.

Key words: data center, energy consumption optimization, load, eXtreme Gradient Boosting (XGBoost), Multiple Gated Recurrent Units (Multi-GRU)

中图分类号:

TP391.9

申明尧, 韩萌, 杜诗语, 孙蕊, 张春砚. 融合XGBoost和Multi-GRU的数据中心服务器能耗优化算法[J]. 计算机应用, 2022, 42(1): 198-208.

Mingyao SHEN, Meng HAN, Shiyu DU, Rui SUN, Chunyan ZHANG. Data center server energy consumption optimization algorithm combining XGBoost and Multi-GRU[J]. Journal of Computer Applications, 2022, 42(1): 198-208.

图/表 19

表1 数据中心能耗构成

Tab. 1 Composition of data center energy consumption

组件	所占百分比/%
服务器	40
交换机和路由器	10
空调	12
制冷系统	25
供电电源	10
照明	3

图1 GRU模型的结构

Fig. 1 GRU model structure

图2 Multi-GRU网络结构

Fig. 2 Multi-GRU network structure

图3 ECOXG算法整体流程

Fig. 3 Overall flowchart of ECOXG algorithm

表2 XGBoost参数

Tab. 2 XGBoost parameters

超参数	取值
n_estimators	20.00
learning_rate	0.30
max_depth	3.00
min_child_weight	2.00
reg_alpha	0.05
reg_lambda	0.60

表3 服务器节点配置参数

Tab. 3 Server node configuration parameters

硬件环境	配置
处理器	Intel XeonE5 2609V3
内存	8 GB DDR4 ECC
硬盘	4 TB SATA
网卡	I210-i

表4 能耗预测模型实验结果

Tab. 4 Experimental results of energy consumption prediction models

模型	RMSE	MAE
XGBoost	8.25	6.65
SVR	23.28	18.11
MLP	25.56	18.57
CNN	11.63	14.59

图4 能耗预测值与真实值的对比

Fig. 4 Comparison of predicted and true values of energy consumption

图5 绝对值误差散点图

Fig. 5 Scatter plot of absolute error

表5 不同GRU网络层的性能

Tab. 5 Performance of different GRU network layers

层数	迭代次数	RMSE	训练时间/s	预测时间/s
1	100	7.69	28.81	34
2	100	7.39	55.47	53
3	100	7.27	84.39	67
4	100	7.35	112.99	88
5	100	7.99	148.53	118
6	100	8.34	172.24	154

图6 不同GRU网络层的RMSE损失变化

Fig. 6 RMSE loss change of different GRU network layers

图7 Batchsize和学习率对模型损失RMSE的影响

Fig. 7 Influence of Batchsize and learning_rate on model loss RMSE

表6 负载模型RMSE结果对比

Tab. 6 Comparison of load model RMSE results

模型	RMSE
Multi-GRU	7.27
LSTM	10.54
CNN-LSTM	9.43
CNN-GRU	10.81
CNN	14.82
MLP	31.38
Transformer	27.99

表7 负载模型效率对比

Tab. 7 Comparison of load model efficiency

模型	训练时间/s	预测时间/s
Multi-GRU	84.39	67.0
LSTM	148.63	85.0
CNN-LSTM	210.46	93.0
CNN-GRU	159.52	75.0
CNN	2.16	1.0
MLP	1.55	0.6
Transformer	1.83	1.0

图8 部分CPU负载预测效果

Fig. 8 Part of CPU load prediction effect

图9 部分内存负载预测效果

Fig. 9 Part of memory load prediction effect

图10 部分网络负载预测效果

Fig. 10 Part of network load prediction effect

图11 部分磁盘负载预测效果

Fig. 11 Part of disk load prediction effect

图12 不同负载时降频前后能耗对比

Fig. 12 Comparison of energy consumption before and after frequency reduction under different loads

参考文献 39

1	AL-TARAZI M， CHANG J M. Performance-aware energy saving for data center networks［J］. IEEE Transactions on Network and Service Management， 2019， 16（1）： 206-219. 10.1109/tnsm.2019.2891826
2	中国制冷学会数据中心冷却工作组. 中国数据中心冷却技术年度发展研究报告2016［M］. 北京：中国建筑工业出版社， 2016： 1-196. 10.18356/9b3b9062-en
	Data Center Cooling Working Group of Chinese Refrigeration Society. Annual Cooling Technology Development Research Report of China Data Center 2016［M］. Beijing： China Architecture and Building Press， 2016： 1-196. 10.18356/9b3b9062-en
3	陈东虎. 基于排队论的云数据中心节能管理策略［D］. 重庆：重庆邮电大学， 2020： 1-58.
	CHEN D H. The energy saving management strategy of cloud data center based on queuing［D］. Chongqing： Chongqing University of Posts and Telecommunications， 2020： 1-58.
4	HELLER B， SEETHARAMAN S， MAHADEVAN P， et al. ElasticTree： saving energy in data center networks［C］// Proceedings of the 7th USENIX Conference on Networked Systems Design and Implementation. Berkeley： USENIX Association， 2010： 249-264.
5	TU R L， WANG X， YANG Y， et al. Energy-saving model for SDN data centers［J］. The Journal of Supercomputing， 2014， 70（3）： 1477-1495. 10.1007/s11227-014-1237-3
6	JING S Y， ALI S， SHE K， et al. State-of-the-art research study for green cloud computing［J］. The Journal of Supercomputing， 2013， 65（1）： 445-468. 10.1007/s11227-011-0722-1
7	LIN W W， WANG H Y， ZHANG Y F， et al. A cloud server energy consumption measurement system for heterogeneous cloud environments［J］. Information Sciences， 2018， 468： 47-62. 10.1016/j.ins.2018.08.032
8	LIANG Y， HU Z G. Prediction method of energy consumption based on multiple energy-related features in data center［C］// Proceedings of the 2019 IEEE International Conference on Parallel and Distributed Processing with Applications， Big Data and Cloud Computing， Sustainable Computing and Communications， Social Computing and Networking. Piscataway： IEEE， 2019： 140-146. 10.1109/ispa-bdcloud-sustaincom-socialcom48970.2019.00030
9	DIOUANI S， MEDROMI H. Energy consumption modeling and prediction in the cloud data centers［J］. Journal of Engineering Science and Technology Review， 2020， 13（3）： 224-234. 10.25103/jestr.133.25
10	STAMATESCU I， PLOIX S， FĂGĂRĂŞAN I， et al. Data center server energy consumption optimization algorithm［C］// Proceedings of the 26th Mediterranean Conference on Control and Automation. Piscataway： IEEE， 2018：813-818. 10.1109/med.2018.8442890
11	YADAV R， ZHANG W Z， LI K Q， et al. An adaptive heuristic for managing energy consumption and overloaded hosts in a cloud data center［J］. Wireless Networks， 2020， 26（3）： 1905-1919. 10.1007/s11276-018-1874-1
12	MAHJOUB Y A E， FOURNEAU J M， CASTEL-TALEB H. Analysis of energy consumption in cloud center with tasks migrations［C］// Proceedings of the 2019 International Conference on Computer Networks， CCIS1039. Cham： Springer， 2019： 301-315.
13	DUOLIKUN D， WATANABE R， ENOKIDO T， et al. A model for migration of virtual machines to reduce electric energy consumption［C］// Proceedings of the 10th International Conference on Complex， Intelligent， and Software Intensive Systems. Piscataway： IEEE， 2016： 159-166. 10.1109/cisis.2016.121
14	SONG Y， ZHAO X， WANG B， et al. Trading-off computing and cooling energies by VM migration in data centers［J］. IEICE Transactions on Information and Systems， 2018， E101.D（9）： 2224-2234. 10.1587/transinf.2017edp7329
15	CHOI J. Virtual machine placement algorithm for energy saving and reliability of servers in cloud data centers［J］. Journal of Network and Systems Management， 2019， 27（1）： 149-165. 10.1007/s10922-018-9462-3
16	CHO K， MERRIËNBOER B VAN， GU̇LÇEHRE Ç， et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation［C］// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Stroudsburg， PA： Association for Computational Linguistics， 2014： 1724-1734. 10.3115/v1/d14-1179
17	CHEN T Q， GUESTRIN C. XGBoost： a scalable tree boosting system［C］// Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2016： 785-794. 10.1145/2939672.2939785
18	ARCE J M M， MACABEBE E Q B. Real-time power consumption monitoring and forecasting using regression techniques and machine learning algorithms［C］// Proceedings of the 2019 IEEE International Conference on Internet of Things and Intelligence System. Piscataway： IEEE， 2019： 135-140. 10.1109/iotais47347.2019.8980380
19	YAN L， MENG L. A simplified prediction model for energy use of air conditioner in residential buildings based on monitoring data from the cloud platform［J］. Sustainable Cities and Society， 2020， 60： No.102194. 10.1016/j.scs.2020.102194
20	MOHAMMADI M， RASHID T A， KARIM S H T， et al. A comprehensive survey and taxonomy of the SVM-based intrusion detection systems［J］. Journal of Network and Computer Applications， 2021， 178： No.102983. 10.1016/j.jnca.2021.102983
21	MA Z T， YE C T， MA W B. Support vector regression for predicting building energy consumption in southern China［J］. Energy Procedia， 2019， 158： 3433-3438. 10.1016/j.egypro.2019.01.931
22	GUO N， GUI W F， CHEN W， et al. Using improved support vector regression to predict the transmitted energy consumption data by distributed wireless sensor network［J］. EURASIP Journal on Wireless Communications and Networking， 2020， 2020： No.120. 10.1186/s13638-020-01729-x
23	LIU D D， YANG Q Q， YANG F. Predicting building energy consumption by time series model based on machine learning and empirical mode decomposition［C］// Proceedings of the 5th IEEE International Conference on Big Data Analytics. Piscataway： IEEE， 2020： 145-150. 10.1109/icbda49040.2020.9101335
24	GOU G H. SVR-based prediction of carbon emissions from energy consumption in Henan Province［J］. IOP Conference Series： Earth and Environmental Science， 2018， 113： No.012052. 10.1088/1755-1315/113/1/012052
25	HUANG Z K， YANG C H， ZHOU X J， et al. Energy consumption forecasting for the nonferrous metallurgy industry using hybrid support vector regression with an adaptive state transition algorithm［J］. Cognitive Computation， 2019， 12（2）： 357-368. 10.1007/s12559-019-09644-0
26	ILBEIGI M， GHOMEISHI M， DEHGHANBANADAKI A. Prediction and optimization of energy consumption in an office building using artificial neural network and a genetic algorithm［J］. Sustainable Cities and Society， 2020， 61： No.102325. 10.1016/j.scs.2020.102325
27	VERGARA G， CÓZAR J， ROMERO-GONZÁLEZ C， et al. Comparing ELM against MLP for electrical power prediction in buildings［C］// Proceedings of the 2015 International Work-Conference on the Interplay Between Natural and Artificial Computation， LNCS9108. Cham： Springer， 2015： 409-418.
28	ZENG C L， WU C C， ZUO L L， et al. Predicting energy consumption of multiproduct pipeline using artificial neural networks［J］. Energy， 2014， 66： 791-798. 10.1016/j.energy.2014.01.062
29	DAI Z W， WANG Y M， WU Y F. Life prediction method of hydrogen energy battery based on MLP and LOESS［C］// Proceedings of the 2020 International Conference on Computer Engineering and Networks， AISC1274. Singapore： Springer， 2021： 551-562.
30	LIU Y X， DONG J， ZUO D C， et al. Experimental analysis and comparison of load prediction algorithms in cloud data center［C］// Proceedings of the IEEE 19th International Conference on Software Quality， Reliability and Security. Piscataway： IEEE， 2019： 197-203. 10.1109/qrs.2019.00036
31	KOLLIA I， KOLLIAS S. A deep learning approach for load demand forecasting of power systems ［C］// Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence. Piscataway： IEEE， 2018： 912-919. 10.1109/ssci.2018.8628644
32	ZHENG J X， CHEN X Y， YU K， et al. Short-term power load forecasting of residential community based on GRU neural network［C］// Proceedings of the 2018 International Conference on Power System Technology. Piscataway： IEEE， 2018： 4862-4868. 10.1109/powercon.2018.8601718
33	BOUKTIF S， FIAZ A， OUNI A， et al. Multi-sequence LSTM-RNN deep learning and metaheuristics for electric load forecasting［J］. Energies， 2020， 13（2）： No.391. 10.3390/en13020391
34	INTEHA A， NAHID-Al-MASOOD. A GRU-GA hybrid model based technique for short term electrical load forecasting［C］// Proceedings of the 2nd International Conference on Robotics， Electrical and Signal Processing Techniques. Piscataway： IEEE， 2021： 515-519. 10.1109/icrest51555.2021.9331156
35	陆继翔，张琪培，杨志宏，等. 基于CNN-LSTM混合神经网络模型的短期负荷预测方法［J］. 电力系统自动化， 2019， 43（8）： 131-137. 10.7500/AEPS20181012004
	LU J X， ZHANG Q P， YANG Z H， et al. Short-term load forecasting method based on CNN-LSTM hybrid neural network model［J］. Automation of Electric Power Systems， 2019， 43（8）： 131-137. 10.7500/AEPS20181012004
36	赵兵，王增平，纪维佳，等. 基于注意力机制的CNN-GRU短期电力负荷预测方法［J］. 电网技术， 2019， 43（12）： 4370-4376.
	ZHAO B， WANG Z P， JI W J， et al. A short-term power load forecasting method based on attention mechanism of CNN-GRU［J］. Power System Technology， 2019， 43（12）： 4370-4376.
37	BRINGYE Z， SIMA D， KOZLOVSZKY M. Power consumption aware big.LITTLE scheduler for Linux operating system［C］// Proceedings of the 2019 IEEE International Work Conference on Bioinspired Intelligence. Piscataway： IEEE， 2019： 139-144. 10.1109/iwobi47054.2019.9114403
38	WANG Y S， LIAO W L， CHANG Y Q. Gated recurrent unit network-based short-term photovoltaic forecasting［J］. Energies， 2018， 11（8）： No.2163. 10.3390/en11082163
39	CHO K， MERRIËNBOER B VAN， BAHDANAU D， et al. On the properties of neural machine translation： encoder-decoder approaches［C］// Proceedings of the 8th Workshop on Syntax， Semantics and Structure in Statistical Translation. Stroudsburg， PA： Association for Computational Linguistics， 2014： 103-111. 10.3115/v1/w14-4012

[1]	赵全, 汤小春, 朱紫钰, 毛安琪, 李战怀. 大规模短时间任务的低延迟集群调度框架[J]. 计算机应用, 2021, 41(8): 2396-2405.
[2]	尹飞, 龙玲莉, 孔峥, 邵涵, 李鑫, 钱柱中. 面向动态负载的集群容器部署方法[J]. 计算机应用, 2021, 41(6): 1581-1588.
[3]	杨翎, 姜春茂. 基于三支决策的虚拟机节能迁移策略[J]. 计算机应用, 2021, 41(4): 990-998.
[4]	许红亮, 杨桂芹, 蒋占军. 基于软件定义网络的数据中心自适应多路径负载均衡算法[J]. 计算机应用, 2021, 41(4): 1160-1164.
[5]	沈忱, 邰凌翔, 彭煜玮. 面向自动参数调优的动态负载匹配方法[J]. 计算机应用, 2021, 41(3): 657-661.
[6]	陈朗, 王让定, 严迪群, 林昱臻. 融合残差网络和极限梯度提升的音频隐写检测模型[J]. 计算机应用, 2021, 41(2): 449-455.
[7]	苟子安, 张晓, 吴东南, 王艳秋. 分布式存储系统中的日志分析与负载特征提取[J]. 计算机应用, 2020, 40(9): 2586-2593.
[8]	蔡豪, 袁正道. 云数据中心基于贪心模式的虚拟机选择算法[J]. 计算机应用, 2020, 40(6): 1707-1713.
[9]	李翠, 陈庆奎. 基于DPDK并行通信的动态监控模型[J]. 《计算机应用》唯一官方网站, 2020, 40(2): 335-341.
[10]	郭茂祖, 张彬, 赵玲玲, 张昱. 基于联合特征和XGBoost的活动语义识别方法[J]. 计算机应用, 2020, 40(11): 3159-3165.
[11]	黄光球, 谢蓉. 考虑节点过载的碳排放空间关联系统级联失效模型[J]. 计算机应用, 2019, 39(6): 1829-1835.
[12]	李启锐, 彭志平, 崔得龙, 何杰光. 容器云环境虚拟资源配置策略的优化[J]. 计算机应用, 2019, 39(3): 784-789.
[13]	李祝红, 赵灿明, 闫龙, 张信明. 智能电网中电力线通信网络负载均衡的机会路由协议[J]. 计算机应用, 2019, 39(3): 812-816.
[14]	蔺凯青, 李志华, 郭曙杰, 李双俐. 云环境中多目标优化的虚拟机放置算法[J]. 计算机应用, 2019, 39(12): 3597-3603.
[15]	刘开南. 云数据中心基于贪心算法的虚拟机迁移策略[J]. 计算机应用, 2019, 39(11): 3333-3338.

融合XGBoost和Multi-GRU的数据中心服务器能耗优化算法

Data center server energy consumption optimization algorithm combining XGBoost and Multi-GRU

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 19

参考文献 39

相关文章 15

编辑推荐

Metrics