Energy efficient distributed computing systems

The energy consumption issue in distributed computing systems raises various monetary, environmental and system performance concerns. Electricity consumption in the US doubled from 2000 to 2005. From a financial and environmental standpoint, reducing the consumption of electricity is important, yet these reforms must not lead to performance degradation of the computing systems. These contradicting constraints create a suite of complex problems that need to be resolved in order to lead to 'greener' distributed computing systems. This book brings together a group of outstanding researchers that investigate the different facets of green and energy efficient distributed computing.

Key features:

One of the first books of its kind Features latest research findings on emerging topics by well-known scientists Valuable research for grad students, postdocs, and researchers Research will greatly feed into other technologies and application domains

Author Notes

ALBERT Y. ZOMAYA is the Chair Professor of High Performance Computing & Networking in the School of Information Technologies, The University of Sydney. He is a Fellow of the IEEE, the American Association for the Advancement of Science, and the Institution of Engineering and Technology, and a Distinguished Engineer of the ACM. He has authored seven books and some 400 articles in technical journals.

YOUNG CHOON LEE, PhD, is with the Centre for Distributed and High Performance Computing, School of Information Technologies, The University of Sydney.

Keqin LiRong Ge and Kirk W. CameronIvan Rodero and Manish ParasharPeng Rong and Massoud PedramAnne-Ce´ cile Orgerie and Laurent Lefe` vre

Preface	p. xxix
Acknowledgments	p. xxxi
Contributors	p. xxxiii
1 Power Allocation and Task Scheduling on Multiprocessor Computers with Energy and Time Constraints	p. 1
1.1 Introduction	p. 1
1.1.1 Energy Consumption	p. 1
1.1.2 Power Reduction	p. 2
1.1.3 Dynamic Power Management	p. 3
1.1.4 Task Scheduling with Energy and Time Constraints	p. 4
1.1.5 Chapter Outline	p. 5
1.2 Preliminaries	p. 5
1.2.1 Power Consumption Model	p. 5
1.2.2 Problem Definitions	p. 6
1.2.3 Task Models	p. 7
1.2.4 Processor Models	p. 8
1.2.5 Scheduling Models	p. 9
1.2.6 Problem Decomposition	p. 9
1.2.7 Types of Algorithms	p. 10
1.3 Problem Analysis	p. 10
1.3.1 Schedule Length Minimization	p. 10
1.3.1.1 Uniprocessor computers	p. 10
1.3.1.2 Multiprocessor computers	p. 11
1.3.2 Energy Consumption Minimization	p. 12
1.3.2.1 Uniprocessor computers	p. 12
1.3.2.2 Multiprocessor computers	p. 13
1.3.3 Strong NP-Hardness	p. 14
1.3.4 Lower Bounds	p. 14
1.3.5 Energy-Delay Trade-off	p. 15
1.4 Pre-Power-Determination Algorithms	p. 16
1.4.1 Overview	p. 16
1.4.2 Performance Measures	p. 17
1.4.3 Equal-Time Algorithms and Analysis	p. 18
1.4.3.1 Schedule length minimization	p. 18
1.4.3.2 Energy consumption minimization	p. 19
1.4.4 Equal-Energy Algorithms and Analysis	p. 19
1.4.4.1 Schedule length minimization	p. 19
1.4.4.2 Energy consumption minimization	p. 21
1.4.5 Equal-Speed Algorithms and Analysis	p. 22
1.4.5.1 Schedule length minimization	p. 22
1.4.5.2 Energy consumption minimization	p. 23
1.4.6 Numerical Data	p. 24
1.4.7 Simulation Results	p. 25
1.5 Post-Power-Determination Algorithms	p. 28
1.5.1 Overview	p. 28
1.5.2 Analysis of List Scheduling Algorithms	p. 29
1.5.2.1 Analysis of algorithm LS	p. 29
1.5.2.2 Analysis of algorithm LRF	p. 30
1.5.3 Application to Schedule Length Minimization	p. 30
1.5.4 Application to Energy Consumption Minimization	p. 31
1.5.5 Numerical Data	p. 32
1.5.6 Simulation Results	p. 32
1.6 Summary and Further Research	p. 33
References	p. 34
2 Power-Aware High Performance Computing	p. 39
2.1 Introduction	p. 39
2.2 Background	p. 41
2.2.1 Current Hardware Technology and Power Consumption	p. 41
2.2.1.1 Processor power	p. 41
2.2.1.2 Memory subsystem power	p. 42
2.2.2 Performance	p. 43
2.2.3 Energy Efficiency	p. 44
2.3 Related Work	p. 45
2.3.1 Power Profiling	p. 45
2.3.1.1 Simulator-based power estimation	p. 45
2.3.1.2 Direct measurements	p. 46
2.3.1.3 Event-based estimation	p. 46
2.3.2 Performance Scalability on Power-Aware Systems	p. 46
2.3.3 Adaptive Power Allocation for Energy-Efficient Computing	p. 47
2.4 PowerPack: Fine-Grain Energy Profiling of HPC Applications	p. 48
2.4.1 Design and Implementation of PowerPack	p. 48
2.4.1.1 Overview	p. 48
2.4.1.2 Fine-grain systematic power measurement	p. 50
2.4.1.3 Automatic power profiling and code synchronization	p. 51
2.4.2 Power Profiles of HPC Applications and Systems	p. 53
2.4.2.1 Power distribution over components	p. 53
2.4.2.2 Power dynamics of applications	p. 54
2.4.2.3 Power bounds on HPC systems	p. 55
2.4.2.4 Power versus dynamic voltage and frequency scaling	p. 57
2.5 Power-Aware Speedup Model	p. 59
2.5.1 Power-Aware Speedup	p. 59
2.5.1.1 Sequential execution time for a single workload T1(w, f )	p. 60
2.5.1.2 Sequential execution time for an ON-chip/OFF-chip workload	p. 60
2.5.1.3 Parallel execution time on N processors for an ON-/OFF-chip workload with DOP = i	p. 61
2.5.1.4 Power-aware speedup for DOP and ON-/OFF-chip workloads	p. 62
2.5.2 Model Parametrization and Validation	p. 63
2.5.2.1 Coarse-grain parametrization and validation	p. 64
2.5.2.2 Fine-grain parametrization and validation	p. 66
2.6 Model Usages	p. 69
2.6.1 Identification of Optimal System Configurations	p. 70
2.6.2 PAS-Directed Energy-Driven Runtime Frequency Scaling	p. 71
2.7 Conclusion	p. 73
References	p. 75
3 Energy Efficiency in HPC Systems	p. 81
3.1 Introduction	p. 81
3.2 Background and Related Work	p. 83
3.2.1 CPU Power Management	p. 83
3.2.1.1 OS-level CPU power management	p. 83
3.2.1.2 Workload-level CPU power management	p. 84
3.2.1.3 Cluster-level CPU power management	p. 84
3.2.2 Component-Based Power Management	p. 85
3.2.2.1 Memory subsystem	p. 85
3.2.2.2 Storage subsystem	p. 86
3.2.3 Thermal-Conscious Power Management	p. 87
3.2.4 Power Management in Virtualized Datacenters	p. 87
3.3 Proactive, Component-Based Power Management	p. 88
3.3.1 Job Allocation Policies	p. 88
3.3.2 Workload Profiling	p. 90
3.4 Quantifying Energy Saving Possibilities	p. 91
3.4.1 Methodology	p. 92
3.4.2 Component-Level Power Requirements	p. 92
3.4.3 Energy Savings	p. 94
3.5 Evaluation of the Proposed Strategies	p. 95
3.5.1 Methodology	p. 96
3.5.2 Workloads	p. 96
3.5.3 Metrics	p. 97
3.6 Results	p. 97
3.7 Concluding Remarks	p. 102
3.8 Summary	p. 103
References	p. 104
4 A Stochastic Framework for Hierarchical System-Level Power Management	p. 109
4.1 Introduction	p. 109
4.2 Related Work	p. 111
4.3 A Hierarchical DPM Architecture	p. 113
4.4 Modeling	p. 114
4.4.1 Model of the Application Pool	p. 114
4.4.2 Model of the Service Flow Control	p. 118
4.4.3 Model of the Simulated Service Provider	p. 119
4.4.4 Modeling Dependencies between SPs	p. 120
4.5 Policy Optimization	p. 122
4.5.1 Mathematical Formulation	p. 122
4.5.2 Optimal Time-Out Policy for Local Power Manager	p. 123
4.6 Experimental Results	p. 125
4.7 Conclusion	p. 130
References	p. 130
5 Energy-Efficient Reservation Infrastructure for Grids, Clouds, and Networks	p. 133
5.1 Introduction	p. 133
5.2 Related Works	p. 134
5.2.1 Server and Data Center Power Management	p. 135
5.2.2 Node Optimizations	p. 135
5.2.3 Virtualization to Improve Energy Efficiency	p. 136
5.2.4 Energy Awareness in Wired Networking Equipment	p. 136
5.2.5 Synthesis	p. 137
5.3 ERIDIS: Energy-Efficient Reservation Infrastructure for Large-Scale Distributed Systems	p. 138
5.3.1 ERIDIS Architecture	p. 138
5.3.2 Management of the Resource Reservations	p. 141
5.3.3 Resource Management and On/Off Algorithms	p. 145
5.3.4 Energy-Consumption Estimates	p. 146
5.3.5 Prediction Algorithms	p. 146
5.4 EARI: Energy-Aware Reservation Infrastructure for Data Centers and Grids	p. 147
5.4.1 EARIÆs Architecture	p. 147
Validation of EARI on Experimental Grid Traces	p. 147
GOC: Green Open Cloud	p. 149
GOCÆs Resource Manager Architecture	p. 150
Validation of the GOC Framework	p. 152
HERMES: High Level Energy-Aware Model for Bandwidth Reservation in End-To-End Networks	p. 152
HERMESÆ Architecture	p. 154
The Reservation Process of HERMES	p. 155
Discussion	p. 157
Summary	p. 158
Problem and Motivation	p. 163
Context	p. 163
Chapter Roadmap	p. 164
Energy-Aware Infrastructures	p. 164
Buildings	p. 165
Context-Aware Buildings	p. 165
Cooling	p. 166
Current Resource Management Practices	p. 167
Widely Used Resource Management Systems	p. 167
Job Requirement Description	p. 169
Scientific and Technical Challenges	p. 170
Theoretical Difficulties	p. 170
Technical Difficulties	p. 170
Controlling and Tuning Jobs	p. 171
Energy-Aware Job Placement Algorithms	p. 172
State of the Art	p. 172
Detailing One Approach	p. 174
Discussion	p. 180
Open Issues and Opportunities	p. 180
Obstacles for Adoption in Production	p. 182
Conclusion	p. 183
Introduction	p. 189
Problem Formulation	p. 191
The System Model	p. 191
PEs	p. 191
DVS	p. 191
Tasks	p. 192
Preliminaries	p. 192
Formulating the Energy-Makespan Minimization Problem	p. 192
Proposed Algorithms	p. 193
Greedy Heuristics	p. 194
Greedy heuristic scheduling algorithm	p. 196
Greedy-min	p. 197
Greedy-deadline	p. 198
Greedy-max	p. 198
MaxMin	p. 199
ObFun	p. 199
MinMin StdDev	p. 202
MinMax StdDev	p. 202
Simulations, Results, and Discussion	p. 203
Workload	p. 203
Comparative Results	p. 204
Small-size problems	p. 204
Large-size problems	p. 206
Related Works	p. 211
Conclusion	p. 211
Heuristic algorithms	p. 219
AI planning	p. 219
Semantic techniques	p. 219
Expert systems and genetic algorithms	p. 220
Instance-based learning	p. 221
Reinforcement learning	p. 222
Feature and example selection	p. 225
Electric power	p. 247
Heat removal	p. 249
Criteria for good metrics	p. 251

Available:*

On Order

Summary

Summary

Author Notes

Table of Contents