小样本负载序列的结构化预测方法

doi:10.3778/j.issn.1673-9418.2101031

计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (7): 1552-1560.DOI: 10.3778/j.issn.1673-9418.2101031

小样本负载序列的结构化预测方法

刘春红¹^,²^,⁺(), 张志华¹, 焦洁¹, 程渤³

1.河南师范大学计算机与信息工程学院,河南新乡 453007
2.智慧商务与物联网技术河南省工程实验室,河南新乡 453007
3.北京邮电大学网络与交换技术国家重点实验室,北京 100876

收稿日期:2020-12-02 修回日期:2021-01-29 出版日期:2022-07-01 发布日期:2021-02-05
作者简介:刘春红（1969—）,女,博士,副教授,主要研究方向为云计算、机器学习、服务计算。
LIU Chunhong, born in 1969, Ph.D., associate professor. Her research interests include cloud com-puting, machine learning and service computing.
张志华（1994—）,女,硕士,主要研究方向为云计算、时间序列预测。
ZHANG Zhihua, born in 1994, M.S. Her research interests include cloud computing and time series prediction.
焦洁（1996—）,女,硕士研究生,CCF学生会员,主要研究方向为云计算、机器学习。
JIAO Jie, born in 1996, M.S. candidate, student member of CCF. Her research interests include cloud computing and machine learning.
程渤（1975—）,男,博士,教授,主要研究方向为服务计算、移动互联网、云计算。
CHENG Bo, born in 1975, Ph.D., professor. His research interests include service compu-ting, mobile Internet and cloud computing.
基金资助:
河南省重点研发与推广专项（科技攻关）项目(202102210163);河南省重点研发与推广专项（科技攻关）项目(202102210152);河南省高等教育教学改革研究与实践项目(2019SJGLX033Y);网络与交换技术国家重点实验室（北京邮电大学）开放课题资助项目(SKLNST-2020-1-02)

Structured Prediction Method for Small Sample Workload Sequences

LIU Chunhong¹^,²^,⁺(), ZHANG Zhihua¹, JIAO Jie¹, CHENG Bo³

1. College of Computer and Information Engineering, Henan Normal University, Xinxiang, Henan 453007, China
2. Engineering Lab of Intelligence Business & Internet of Things, Xinxiang, Henan 453007, China
3. State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China

Received:2020-12-02 Revised:2021-01-29 Online:2022-07-01 Published:2021-02-05
Supported by:
the Key Research and Development and Promotion Special (Technology Research) Project of Henan Province(202102210163);the Key Research and Development and Promotion Special (Technology Research) Project of Henan Province(202102210152);the Research and Practice Project of Higher Education Teaching Reform in Henan Province(2019SJGLX033Y);the Open Foundation of State Key Laboratory of Networking and Switching Technology (Beijing University of Posts and Telecommunications)(SKLNST-2020-1-02)

摘要/Abstract

摘要：

准确的负载预测是实现云平台弹性资源管理的关键。针对云负载预测中存在大量运行周期较短的任务,导致预测模型训练数据不足的问题,提出一种利用多变量负载序列结构化信息的预测方法（SP-MWS）。依据同一任务运行中消耗的多种资源间内在的相关性,挖掘多维负载序列间的信息,补充小样本序列的预测信息量。首先,为获取相关性强的负载类型,采用最大互信息系数（MIC）和信息熵进行负载类型的度量选择;然后,构建核范数正则化多任务学习模型（TNR-MTL）,将相关负载序列同时输入,实现其结构化信息的挖掘,并完成多种负载的同时预测。在Google云平台的运行监控日志数据集上进行验证,实验结果表明,所提方法获得的相关负载序列类型可明显增加模型信息量。对于预测模型的决策依据进行解释性分析,可视化每种变量对预测结果的贡献度;对比实验表明,所提预测方法在时间性能和预测精度上均优于常用的负载预测方法。

关键词: 云计算, 弹性, 负载预测, 多变量预测, 结构化信息

Abstract:

Accurate workload prediction is the key to realize elastic resource management of cloud platform. Aiming at the problem that a large number of tasks with short running time achieve prediction in the cloud platform, which leads to the lack of training data of the forecasting model, a structured prediction of multivariable workload sequences (SP-MWS) method is proposed. It is based on the characteristics of intrinsic correlation among multiple resources consumed in the running of a single task, and the relationship of multi-dimensional workload sequences is explored to supplement the prediction information of small-scale sequence. Firstly, in order to obtain the related workload types, the maximum information coefficient (MIC) and information entropy are adopted to measure the correlation, and related workload types are selected. Secondly, for the selected multiple related workloads, trace-norm regularization multi-task learning (TNR-MTL) is introduced to construct prediction model to realize structural information mining of related workload sequences and complete prediction of multiple workloads simultaneously. Validated on Google cloud platform’s operational monitoring log dataset, the experimental results show that the proposed method can significantly increase model information; the decision-making basis of the prediction model is interpreted and the contribution of each variable to the prediction model is visualized. Comparative experiments show that, the proposed prediction method is better than the commonly used workload prediction methods in time performance and prediction accuracy.

Key words: cloud computing, elasticity, workload prediction, multivariable prediction, structural information

中图分类号:

TP393

刘春红, 张志华, 焦洁, 程渤. 小样本负载序列的结构化预测方法[J]. 计算机科学与探索, 2022, 16(7): 1552-1560.

LIU Chunhong, ZHANG Zhihua, JIAO Jie, CHENG Bo. Structured Prediction Method for Small Sample Workload Sequences[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(7): 1552-1560.

图/表 13

参考文献 17

[1]	AL-DHURAIBI Y, PARAISO F, DJARALLAH N, et al. Ela-sticity in cloud computing: state of the art and research chal-lenges[J]. IEEE Transactions on Services Computing, 2018, 11(2): 430-447. DOI URL
[2]	COUTINHO E F, DE CARVALHO SOUSA F R, REGO P A L, et al. Elasticity in cloud computing: a survey[J]. Anna-les des Télécommunications, 2015, 70(7/8): 289-309.
[3]	CALHEIROS R N, MASOUMI E, RANJAN R, et al. Work-load prediction using ARIMA model and its impact on cloud applications’ QoS[J]. IEEE Transactions on Cloud Compu-ting, 2015, 3(4): 449-458.
[4]	LIU C H, LIU C C, SHANG Y L, et al. An adaptive predic-tion approach based on workload pattern discrimination in the cloud[J]. Journal of Network and Computer Applications, 2017, 80: 35-44. DOI URL
[5]	NGUYEN H M, GAURAV K, DAEYOUNG K. Host load prediction in cloud computing using long short-term memory encoder-decoder[J]. The Journal of Supercomputing, 2019, 75(11): 7592-7605. DOI URL
[6]	HIEU N T, DI FRANCESCO M, YLÄ-JÄÄSKI A. Virtual machine consolidation with multiple usage prediction for energy-efficient cloud data centers[J]. IEEE Transactions on Services Computing, 2020, 13(1): 186-199. DOI URL
[7]	TRAN N, NGUYEN T, NGUYEN B M, et al. A multivariate fuzzy time series resource forecast model for clouds using LSTM and data correlation analysis[C]// Proceedings of the 22nd International Conference KES-2018, Belgrade, Sep 3-5, 2018. New York: Elsevier Science, 2018: 636-645.
[8]	ISLAM S, KEUNG J, LEE K, et al. Empirical prediction mo-dels for adaptive resource provisioning in the cloud[J]. Fu-ture Generation Computer Systems, 2012, 28(1): 155-162.
[9]	REISS C, TUMANOV A, GANGER G R, et al. Heterogeneity and dynamicity of clouds at scale: Google trace analysis[C]// Proceedings of the 2012 ACM Symposium on Cloud Compu-ting, San Jose, Oct 14-17, 2012. New York: ACM, 2012: 1-13.
[10]	VERGARA J R, ESTEVEZ P. A review of feature selec-tion methods based on mutual information[J]. Neural Com-puting & Applications, 2014, 24(1): 175-186.
[11]	张俐, 袁玉宇, 王枞. 基于最大相关信息系数的FCBF特征选择算法[J]. 北京邮电大学学报, 2018, 41(4): 86-90.
	ZHANG L, YUAN Y Y, WANG C. FCBF feature selection algorithm based on maximum information coefficient[J]. Journal of Beijing University of Posts and Telecom-munications, 2018, 41(4): 86-90.
[12]	殷怡. 复杂时间序列的相关性及信息熵研究[D]. 北京: 北京交通大学, 2018.
	YIN Y. Study on the correlation and information entropy of complex time series[D]. Beijing: Beijing Jiaotong Univer-sity, 2018.
[13]	王二威, 瞿英, 吴祈宗. 基于信息熵的软件可靠性多模型动态预测方法[J]. 桂林理工大学学报, 2016, 36(3): 603-608.
	WANG E W, QU Y, WU Q Z. Multi-model dynamic predic-tion method of software reliability based on information entropy[J]. Journal of Guilin University of Technology, 2016, 36(3): 603-608.
[14]	REISS C, WILKES J, HELLERSTEIN J L. Google cluster-usage traces: format+schema[Z]. Google Inc., 2011: 1-14.
[15]	DOU C X, BIAN Z W, ZHANG T F. Short-term photovo-ltaic output forecasting based on multivariable phase space reconstruction and support vector regression[C]// Proceedings of the 5th IEEE International Conference on Cloud Com-puting and Intelligence Systems, Nanjing, Nov 23-25, 2018. Piscataway: IEEE, 2018: 983-987.
[16]	RIBEIRO M T, SINGH S, GUESTRIN C. “Why should I trust you?”: explaining the predictions of any classifier[C]// Proceedings of the Demonstrations Session, the 2016 Confe-rence of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, San Diego, Jun 12-17, 2016. Stroudsburg: ACL, 2016: 1135-1144.
[17]	STRUMBELJ E, KONONENKO I. Explaining prediction models and individual predictions with feature contributions[J]. Knowledge & Information Systems, 2014, 41(3): 647-665.

待预测负载类型	其他负载类型（从大到小）
CPU	CPU、内存、meanDiskIO、aMem、maxDiskIO、maxCPU、unPCM、maxMem、toPCM、磁盘
内存	内存、CPU、meanDiskIO、aMem、maxMem、toPCM、unPCM、maxDiskIO、maxCPU、磁盘
磁盘	磁盘、CPU、aMem、内存、meanDiskIO、maxDiskIO、maxCPU、unPCM、maxMem、toPCM

待预测负载类型	其他负载类型（从大到小）
CPU	CPU、内存、meanDiskIO、aMem、maxDiskIO、maxCPU、unPCM、maxMem、toPCM、磁盘
内存	内存、CPU、meanDiskIO、aMem、maxMem、toPCM、unPCM、maxDiskIO、maxCPU、磁盘
磁盘	磁盘、CPU、aMem、内存、meanDiskIO、maxDiskIO、maxCPU、unPCM、maxMem、toPCM

阶段	RMSE	MAE	MRE	SMAPE
V3	6.403 8E-04	4.970 9E-04	0.015 4	0.007 7
V4	6.515 2E-04	4.812 9E-04	0.015 0	0.007 5
V5	6.229 2E-04	4.529 1E-04	0.014 1	0.007 1
V6	5.298 3E-04	3.750 1E-04	0.011 7	0.005 9

阶段	RMSE	MAE	MRE	SMAPE
V3	6.403 8E-04	4.970 9E-04	0.015 4	0.007 7
V4	6.515 2E-04	4.812 9E-04	0.015 0	0.007 5
V5	6.229 2E-04	4.529 1E-04	0.014 1	0.007 1
V6	5.298 3E-04	3.750 1E-04	0.011 7	0.005 9

方法	RMSE	MAE	MRE	SMAPE
SP-MWS	5.136 8E-04	3.695 0E-04	0.011 1	0.005 6
多元线性	0.001 3	7.830 1E-04	0.023 5	0.011 9
P-SVR	0.001 7	0.001 3	0.037 6	0.019 2
M-LSTM	0.002 0	0.001 4	0.040 0	0.020 5

小样本负载序列的结构化预测方法

Structured Prediction Method for Small Sample Workload Sequences

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 17

相关文章 15

编辑推荐

Metrics

方法	RMSE	MAE	MRE	SMAPE
SP-MWS	0.001 5	9.117 0E-04	0.010 9	0.005 4
多元线性	0.003 2	0.001 8	0.020 7	0.010 5
P-SVR	0.001 2	9.252 2E-04	0.011 0	0.005 5
M-LSTM	0.002 4	0.002 0	0.023 6	0.011 8

方法	RMSE	MAE	MRE	SMAPE
SP-MWS	2.233 3E-04	1.368 4E-04	0.012 7	0.006 4
多元线性	6.399 7E-04	4.585 7E-04	0.044 5	0.022 5
P-SVR	0.001 1	7.655 8E-04	0.079 1	0.038 1
M-LSTM	0.002 2	0.001 6	0.169 4	0.084 2

[1]	吴虹佳，刘芳，刘斌，蔡志平. 分散计算：技术、应用与挑战[J]. 计算机科学与探索, 2020, 14(5): 721-730.
[2]	郑良汉，何亨，童潜，杨湘，陈享. 云环境中的多授权机构访问控制方案[J]. 计算机科学与探索, 2020, 14(11): 1865-1878.
[3]	张胜霞，田呈亮. 在幺模矩阵加密方法下的安全外包算法[J]. 计算机科学与探索, 2020, 14(1): 73-82.
[4]	陈彦橦，裴树军，苗辉. 云科学工作流截止期限约束代价优化调度算法[J]. 计算机科学与探索, 2019, 13(8): 1307-1318.
[5]	任晓莉，杨建卫，李乃乾. 云计算中基于动态虚拟化电子流密码的安全存储[J]. 计算机科学与探索, 2019, 13(8): 1331-1340.
[6]	浦建宇，陈蕾，邵楷. 基于Katz增强归纳型矩阵补全的基因-疾病关联关系预测[J]. 计算机科学与探索, 2019, 13(7): 1154-1164.
[7]	赵倩，谢上钦，韩轲，龚青泽，冯光升，林俊宇. 远程直接内存访问与检查点相结合的容器迁移[J]. 计算机科学与探索, 2019, 13(12): 1995-2007.
[8]	吕小敬，刘钊，蒋令闻，陈德训，杨广文. 船舶三维声弹性模拟软件的并行优化策略[J]. 计算机科学与探索, 2019, 13(11): 1852-1863.
[9]	齐小刚，张碧雯，刘立芳，胡绍林. 复杂信息网络的弹性评估和优化方法研究[J]. 计算机科学与探索, 2018, 12(8): 1252-1262.
[10]	李勇，滕飞，黄齐川，李天瑞. 基于Spark的时间序列并行分解模型[J]. 计算机科学与探索, 2018, 12(7): 1055-1063.
[11]	裴树军，宋冬梅，孔德凯. Map/Reduce下快速剪枝算法在复杂任务调度中的应用[J]. 计算机科学与探索, 2018, 12(1): 72-81.
[12]	刘沛东，安博，钟业弘，王虎，曹东刚. 私有云环境下基于虚拟集群的资源共享方法[J]. 计算机科学与探索, 2017, 11(8): 1204-1213.
[13]	詹杭龙，曹东刚，谢冰. 分布共享环境下支持弹性伸缩的图处理框架[J]. 计算机科学与探索, 2016, 10(7): 901-914.
[14]	王虹旭，吴斌，刘旸. 基于Spark的并行图数据分析系统[J]. 计算机科学与探索, 2015, 9(9): 1066-1074.
[15]	宋宝燕，李雪城，任才，丁琳琳. 云环境下分层的中间数据容错方法[J]. 计算机科学与探索, 2015, 9(5): 546-554.