Resource management in virtualized data centers regarding performance and energy aspects [Elektronische Ressource] / vorgelegt von Marko Hoyer

Resource management in virtualized data centers regarding performance and energy aspects [Elektronische Ressource] / vorgelegt von Marko Hoyer

English
199 Pages
Read
Download
Downloading requires you to have access to the YouScribe library
Learn all about the services we offer

Description

Fakultat II { Informatik, Wirtschafts- und RechtswissenschaftenDepartment fur InformatikResource Management in Virtualized DataCenters Regarding Performance andEnergy AspectsDissertation zur Erlangung des Grades einesDoktors der Ingenieurwissenschaftenvorgelegt vonDipl.-Inform. Marko HoyerDatum der Disputation24. Mai, 2011GutachterProf. Dr. Wolfgang NebelProf. Dr. Michael SonnenscheinContents1 Introduction 11.1 Static Resource Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2 Dynamic Resourcet . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Contributions of this Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.4 Document Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 Context and Related Work 72.1 IT Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Data Center Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.3 Operating Systems, IT Services, and Software . . . . . . . . . . . . . . . . . . . 82.3.1 Power Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.3.2 Resourcet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 Problem Statement 133.1 Technical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133.1.1 Service Level Agreements . . . . . . . . . . . . . . . . . . . . . . . . . . 133.1.2 Server Virtualization . . . . . . . . . . . . . . . . . .

Subjects

Informations

Published by
Published 01 January 2011
Reads 8
Language English
Document size 1 MB
Report a problem

Fakultat II { Informatik, Wirtschafts- und Rechtswissenschaften
Department fur Informatik
Resource Management in Virtualized Data
Centers Regarding Performance and
Energy Aspects
Dissertation zur Erlangung des Grades eines
Doktors der Ingenieurwissenschaften
vorgelegt von
Dipl.-Inform. Marko Hoyer
Datum der Disputation
24. Mai, 2011
Gutachter
Prof. Dr. Wolfgang Nebel
Prof. Dr. Michael SonnenscheinContents
1 Introduction 1
1.1 Static Resource Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Dynamic Resourcet . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Contributions of this Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Document Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Context and Related Work 7
2.1 IT Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Data Center Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Operating Systems, IT Services, and Software . . . . . . . . . . . . . . . . . . . 8
2.3.1 Power Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.2 Resourcet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Problem Statement 13
3.1 Technical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.1 Service Level Agreements . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 Server Virtualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.3 Server and Live Migration . . . . . . . . . . . . . . . . . 17
3.1.4 Dealing with Shared Resources in Virtualized Data Centers . . . . . . . 18
3.1.5 Power States of Servers . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Conceptual View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.1 Pessimistic Static Resource Management . . . . . . . . . . . . . . . . . . 23
3.2.2 Optimized Static Management . . . . . . . . . . . . . . . . . . 24
3.2.3 Dynamic Resource Management . . . . . . . . . . . . . . . . . . . . . . 24
3.3 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3.1 Involved Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3.2 Limited Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.3 Overhead and Prerequisites of Control Mechanisms . . . . . . . . . . . . 30
3.3.4 Service Level Agreements . . . . . . . . . . . . . . . . . . . . . . . . . . 32
iiiContents
3.4 Formal De nition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Terminology and Declarations . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4.2 Problem De nition: Static Resource Management . . . . . . . . . . . . 36
3.4.3 Dynamic Resourcet . . . . . . . . . . . 37
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4 Pessimistic Static Resource Management 41
4.1 Service Level Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Modeling the Resource Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 Static Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3.1 Known Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.2 Vector Bin Packing and Resource Management . . . . . . . . . . . . . . 43
5 Statistical Static Resource Management 45
5.1 Mathematical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.1.1 Discrete Random Variables . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.1.2 Operations on Discrete Random Variables . . . . . . . . . . . . . . . . . 46
5.1.3 Stochastic Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.1.4 Probabilities of Realizations of Stochastic Processes . . . . . . . . . . . 49
5.2 Service Level Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2.1 Known Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2.2 Fine Grained SLO Speci cation . . . . . . . . . . . . . . . . . . . . . . . 51
5.2.3 Mapping Performance Metrics on Required Resource Capacity . . . . . 52
5.2.4 Deriving Constraints for Autonomous Resource Management . . . . . . 54
5.2.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Modeling the Resource Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.3.1 Requirements on the Model . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.3.2 Known Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3.3 Modelingh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.4 Static Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.4.1 Known Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.4.2 Pessimistic Statistical Scheduling . . . . . . . . . . . . . . . . . . . . . . 67
5.4.3 Interdependence between Required and Provided Resource Capacity . . 73
5.4.4 Separating Seasonal Trend and Noise from Long Term Trend . . . . . . 75
5.4.5 Using Correlations for Improved Statistical Scheduling . . . . . . . . . . 76
5.4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
ivContents
5.5 Changes in Demand Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.5.1 Impact of Changed Demand Behavior . . . . . . . . . . . . . . . . . . . 81
5.5.2 Detecting Behavior . . . . . . . . . . . . . . . . . . . 82
5.5.3 Preventing SLO Violations Caused by Changed Demand Behavior . . . 82
5.5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6 Dynamic Resource Management 85
6.1 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.1.1 Autocorrelation Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.1.2 Testing Whether a Graph is Acyclic . . . . . . . . . . . . . . . . . . . . 86
6.2 Service Level Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.3 Modeling the Resource Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.3.1 Requirements on the Model . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.3.2 Known Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.3.3 Modeling Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.4 Dynamic Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4.1 Known Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4.2 Basic Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.4.3 Ensuring Resource Constraints . . . . . . . . . . . . . . . . . . . . . . . 100
6.4.4 Extracting a Set of Feasible Operations . . . . . . . . . . . . . . . . . . 103
6.4.5 Ensuring Time Constraints . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.4.6 Scheduling Algorithm - Overview . . . . . . . . . . . . . . . . . . . . . . 108
6.4.7 Scheduling - Consolidating VMs . . . . . . . . . . . . . . . . 109
6.4.8 Scheduling Algorithm - Resolving Resource Shortages . . . . . . . . . . 110
6.4.9 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.5 Changes in Demand Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.5.1 Impact of Changed Demand Behavior . . . . . . . . . . . . . . . . . . . 121
6.5.2 Detecting Behavior . . . . . . . . . . . . . . . . . . . 121
6.5.3 Adapting the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.5.4 Resolving Resource Shortages . . . . . . . . . . . . . . . . . . . . . . . . 123
6.5.5 Limiting the Impact of Changed Demand Behavior . . . . . . . . . . . . 123
6.5.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
vContents
7 Experimental Assessment 127
7.1 Fine Grained QoS Speci cation . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.1.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.1.2 Comparison to Known Approaches . . . . . . . . . . . . . . . . . . . . . 131
7.1.3 In uence of the Number of De ned Performance Goals . . . . . . . . . . 132
7.1.4 Conclusion and Limits of the Analyses . . . . . . . . . . . . . . . . . . . 133
7.2 Resource Demand Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.2.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.2.2 Comparison to Known Approaches . . . . . . . . . . . . . . . . . . . . . 136
7.2.3 Finding the Predominant Period . . . . . . . . . . . . . . . . . . . . . . 139
7.2.4 In uence of Minimal Duration of Saving Intervals . . . . . . . . . . . . . 140
7.2.5 of Long Term Trends . . . . . . . . . . . . . . . . . . . . . . . 141
7.2.6 Di erent VMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.2.7 Conclusion and Limits of the Analyses . . . . . . . . . . . . . . . . . . . 143
7.3 Statistical Static Resource Management . . . . . . . . . . . . . . . . . . . . . . 144
7.3.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.3.2 Comparison to Known Approaches . . . . . . . . . . . . . . . . . . . . . 146
7.3.3 In uence of Server Con guration . . . . . . . . . . . . . . . . . . . . . . 149
7.3.4 Expected Power Savings in Data Centers . . . . . . . . . . . . . . . . . 150
7.3.5 Conclusion and Limits of the Analyses . . . . . . . . . . . . . . . . . . . 151
7.4 Dynamic Resource Management . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.4.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
7.4.2 Comparison to Known Approaches . . . . . . . . . . . . . . . . . . . . . 154
7.4.3 In uence of Server Con guration and Virtualization Environment . . . 155
7.4.4 Limiting the Impact of Forecasting Errors . . . . . . . . . . . . . . . . . 156
7.4.5 Scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.4.6 Conclusion and Limits of Analyses . . . . . . . . . . . . . . . . . . . . . 159
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
8 Summary and Conclusion 163
8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Glossary 167
Bibliography 169
viviiAcknowledgement
First of all I would like to thank my supervisor Prof. Dr. Wolfgang Nebel for his support
and helpful advices concerning the principals of academic work. His contacts to industrial
companies further helped me to gather some practical insights into the topic of this thesis.
The theoretical concepts developed in this thesis could very closely address real practical issues
due to this background. In addition, I would like to thank Prof. Dr. Michael Sonnenschein
for taking the time to review this document.
Much of the work presented in this thesis was supported by two of my students: Pierre
Petliczew and Daniel Schlitt. Thank you both for your good work. I would also like to thank
my colleagues for many constructive discussions; especially Henrik Lipskoch for helping me
with mathematical background, Kiril Schroder and Daniel Schlitt for discussions about the
concepts, and Domenik Helms and Gunnar Schomaker for some tips concerning the formal
representation of some optimization problems. Additionally, special thanks to the NOWIS
company for providing me very good evaluation data to assess my concepts.
A large portion of my work was further supported by the OFFIS Institute for Information
Technology. I did most of the research within the research project \Energy E ciency in Data
Centers". Finally, an internal scholarship helped me to nish my work and to write down this
thesis.
And last but not least I want to thank my family for their support. Especially the last two
month of my work were hard for me for several reasons. Thank you for your help.
***
Ein spezieller Dank soll an dieser Stelle an meine Familie fur ihre mentale Unterstutzung
gehen. Die letzten zwei Monate dieser Arbeit waren aus verschiedenen Grunden nicht einfach
fur mich. Vielen Dank fur Eure Hilfe.
ix