27 2 1

28 2 9

27 2 1

1 4 2 6 2.1........................... 6 2.2...................................... 6 2.3...................... 6 2.4.................................. 7 2.4.1.............................. 7 2.4.2............................. 8 2.5................................. 9 2.6................. 9 3 10 3.1...................................... 10 3.2............................ 10 3.3............................... 11 3.3.1............................... 12 3.3.2.............................. 12 3.3.3............................... 12 3.3.4............................... 12 3.3.5........................... 13 3.4........................................ 14 3.5............................... 14 3.6...................................... 15 3.7...................................... 16 3.8............................. 17 3.9...................................... 18 3.10...................................... 20 3.11............................ 22 4 24 4.1......................... 24 4.2................................... 25 2

4.3.................................... 25 4.4.................................. 26 4.5........................................ 26 4.6............................. 26 4.7...................................... 26 4.8........................................ 28 4.8.1.................................. 29 4.9...................................... 29 4.9.1.................................. 31 5 32 33 34 3

1 [32] [31] CPU [5] [7] [2] 2 10 4

2 3 4 5 5

2 2.1 Service Level Agreement( SLA) SLA [34] NTT VPN Arcstar Universal One SLA [33] 35msec 10% 2.2 2 CPU 2.3 6

[7] [2] 2.4 2 2.4.1 ( CL) [13] ( ) [21] 1 CL A B C 3 1: 7

CPU n n 2.4.2 ( BN) E P (E) [14] 3 [5][19] [26] 2 BN B C P (B) P (C) B C A P (A B, C) BN 2: Web CPU 100% 8

BN 2.5 CL BN [16][15][5][19][12] 2.6 [20] BN CL BN BN 9

3 CL BN 3.1 CL CL BN CL BN CL 3.2 0 CL 1 CL 2 3 BN 4 BN 10

2 BN 4 3 3: CL R k-means[9] BN R bnlearn[26] 3.3 4 2 1 11

3.3.1 Apache JMeter[23] Apache JMeter - Java 3.3.2 Apache mod proxy balancer[24] mod proxy balancer Apache 2 3.3.3 Apache Tomcat[25] ibatis JPetStore[30] Apache Tomcat Java Servlet JavaServer Pages HTML Java JPetStore Java 3.3.4 MySQL MySQL Oracle[29] JPetStore 3.3.4 12

4: 4 3 CPU 1 1GB HDD 20GB 1Gbps CPU 2 4GB HDD 110GB 1Gbps OS Linux(CentOS 6.4) 1 1: CPU OS 2core 1GB 30GB Apache JMeter 2.10 1core 1GB 30GB Apache mod proxy balancer A 1core 1GB 30GB Apache Tomcat 6.0.24-57 B 1core 1GB 30GB ibatis JPetStore 4.0.5 1core 4GB 110GB MySQL 5.1.69-1 CentOS Ver6.4 x64 CPU Disk 10 collectd[27] 3.3.5 2 1 1 3 5 4 2 1 4 8 1(2 ) 2, 3, 4, 5 17 13

2: 1 CPU( 2 ) (%) 2 (bytes) 3 (bytes/sec) 4 Disk( ) I/O (ops/sec) Web Access 5 ( ) 3.4 CL BN BN 3.5 BN COR(0 COR 1) 1 x, y ni=1 (x i x)(y i ȳ) COR = ni=1 (x i x) 2 ni=1 (y i ȳ) 2 COR CL COR CL 14

3.6 BN 1 BN CL 15 5 3 3 1 2 3 1-2 2-3 1-3( ) 6 CL 1-2 6 2 BN 3 5 5: 15

3.7 35 5 1-7 JPetStore Apache JMeter stress [30] ( 6) 7 28 BN 7 3 6: 16

7: 3: 0 00-5 00 5 00-10 00 10 00-15 00 15 00-20 00 20 00-25 00 25 00-30 00 30 00-35 00 A A B A B DB B DB DB 3.8 CL BN 17

CL CL 1 CL 500 CL 500 BN BN 3 3 3 3.9 7 8 10 ( sec) 10 ( sec) 10 ( request/sec) ( ) MA=5 5 4 2-6 18

8: (3.2 ) 1 2 1 BN CL 9 BN 3 P( ) CL D( ) (sec) BN 0 3-6 CL 500 3 4 3-6 BN 10 9 19

9: BN CL ( 1-1) BN 8 BN COR = 0.88 3.10 (3.6 ) 28 4 20

10: CL BN ( 3-6) 4: 1 2-7 0.907 6 2 1-7 0.893 7 3 1-6 0.892 6 4 2-6 0.89 5 5 3-7 0.888 5 6 3-6 0.88 4 7 4-6 0.848 3 8 5-5 0.826 1 9 4-5 0.805 2 10 3-5 0.794 3 27 6-7 21 0.138 2 28 1-1 0 1-0.684 3

4 6 7 6 3.11 ( 1-7) BN CL ( 3-6) BN 1-7 BN CL 11 9 10 11: BN CL ( 1-7) ( 10 11) COR = 0.993 22

23

4 CL BN 4.1 BN 0 CL 1 CL BN 2 BN 3 BN 1-3 12 24

12: 4.2 collectd 10 1-3 30 1 1 6 4.3 CL 25

4.4 CL CL 2 1 6 4.5 35 ( 7) 4.6 3.8 CL BN 1 CL 500 (6 4 ) BN 4.7 7 13 10 ( sec) 10 ( sec) 10 ( request/sec) ( ) 4 2-6 BN 15 (sec) BN 3 ( ) (LS) (LW) R bnlearn 26

13: LW LW 14: 27

LW 13 15 14 15: LW 13 4.8 10 28

4.8.1 BN 1 1 1 3 1 1 3 1 4.9 4 3 BN 10 16 29

16: BN t = 2( ) 1 1 2 4 1 2 2 2 30

4.9.1 5 5: 107 60 90 / 0.561% 0.841% 64 102 57 90 / 0.891 0.882 102 90 10 0 22 48 31

5 10 2 1 2 32

Software Development and Analysis 33

[1] T. F. Abdelzaher, K. G. Shin, et al, Performance guarantees for Web server endsystems: A controltheoretical approach, IEEE Transactions on Parallel and DistributedSystems, 13(1), pp80-96, 2002. [2] G. A. Alvarez, E. Borowsky, et al. An automated resource provisioning tool for largescale storage systems, ACM Transactions on Computer Systems (TOCS), pp483-518, 2001. [3] D. Arthur, B. Manthey, H. Roglin, k-means has polynomial smoothed complexity, 2009. [4] C. Bernard, An optimal convex hull algorithm in any fixed dimension, 1993. [5] I. Cohen, M. Goldszmidt, T. Kelly, J. Symons, J.S. Chase, Correlating instrumentation data to system state: A building block for automated diagnosis and control, USENIX Association OSDI 04: 6th Symposium on Operating Systems Design and Implementation, 2004. [6] U. Fayyad, G. Piatetsky-Shapiro, and P. Smyth From Data Mining to Knowledge Discovery in Databases, 2008. [7] A. Fox and D. Patterson, Self-repairing computers, Scientific American, 2003. [8] N. Friedman, D. Geiger, M. Goldszmidt, Bayesian Network Classifiers, Machine Learning Volume 29, Issue 2-3, pp131-163, 1997. [9] N. Friedman, M. Linial, I. Nachman, D. Pe er, Using Bayesian Networks to Analyze Expression Data, Journal of Computational Biology 7, pp601-620, 2000. [10] S. Iwata, K. Kono, Clustering Performance Anomalies Based on Similarity in Processing Time Changes, IPSJ Transactions on Advanced Computing Systems, Vol.5 No.1 1-12, 2012. [11] J. B. MacQueen, Some Methods for classification and Analysis of Multivariate Observations, Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability,University of California Press, pp281-297, 1967. 34

[12] T.H.D. Nguyen, M. Nagappan, A.E. Hassan, M. Nasser, P. Flora, An Industrial Case Study of Automatically Identifying Performance Regression-Causes, MSR Hyderabad, India, 2014. [13] Y. Okada, T. Sahara, S. Ohgiya, T. Nagashima, Detection of Cluster Boundary in Microarray Data by Reference to MIPS Functional Catalogue Database, The 16th Int. Conference on Genome Informatics, Japanese Society for Bioinformatics, Proc. of The 16th Int. Conference on Genome Informatics, Tokyo, Japan, 2005. [14] J. Pearl, Bayesian Networks, a Model of Self-Activated Memory for Evidential Reasoning, Proceedings, Cognitive Science Society pp329-334, 1985. [15] E. Stehle, K. Lynch, M. Shevertalov, C. Rorres, and S. Mancoridis, On the use of Computational Geometry to Detect Software Faults at Runtime, 7th International Conference on Autonomic Computing, ICAC, Washington, DC, USA, 2010. [16] E. Stehle, K. Lynch, M. Shevertalov, C. Rorres, and S. Mancoridis, Diagnosis of Software Failures Using Computational Geometry, 26th IEEE/ACM International Conference on Automated Software Engineering (ASE 2011), Lawrence, KS, USA, Nov., 2011. [17] R. Taylor, E. Rdcs. Interpretation of the Correlation Coefficient: A Basic Review, JDMS1, pp35-39, 1990. [18] S. Thrun, C. C. Faloutsos, A. W. Moore, P. Spirtes, G. F. Cooper, Learning Bayesian Network Model Structure from Data 2003. [19] S. Zhang, I. Cohen, M. Goldszmidt, J. Symons, A. Fox, Ensembles of Models for Automated Diagnosis of System Performance Problems, The International Conference on Dependable Systems and Networks, Yokohama, Japan, 2005. [20],,,,, Web,, Vol.J99-D, No.1, pp.100-108, 2016. [21],,,,, pp.506-511, 2012. 35

[22],,,,, 188 SE, pp1-8, 2015. [23] Apache JMeter, https://jmeter.apache.org/. [24] Apache mod proxy balancer, http://httpd.apache.org/ docs/2.2/ja/mod/mod proxy balancer.html/. [25] Apache Tomcat, http://tomcat.apache.org/. [26] bnlearn - an R package for Bayesian network learning and inference, http://www.bnlearn.com/. [27] collectd, http://collectd.org/. [28] ibatis JPetStore, http://sourceforge.net/projects/ibatisjpetstore/. [29] Oracle, http://www.oracle.com/. [30] stress project page, http://people.seas.harvard.edu/-apw/stress/. [31],, http://thinkit.co.jp/article/1089/1. [32],, http://www.itmedia.co.jp/ im/articles/1005/19/news106.html. [33] VPN Arcstar Universal One SLA, http://www.ntt.com/vpn/data/sla.html. [34] Web Forum, SLA, http://web-tan.forum.impressrd.jp/g/sla. 36