Energy-Efficient Distributed Computing Systems

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.

• PREFACE xxix ACKNOWLEDGMENTS xxxiCONTRIBUTORS xxxiii1 POWER ALLOCATION AND TASK SCHEDULING ON MULTIPROCESSOR COMPUTERS WITH ENERGY AND TIME CONSTRAINTS 1Keqin Li1.1 Introduction 11.1.1 Energy Consumption 11.1.2 Power Reduction 21.1.3 Dynamic Power Management 31.1.4 Task Scheduling with Energy and Time Constraints 41.1.5 Chapter Outline 51.2 Preliminaries 51.2.1 Power Consumption Model 51.2.2 Problem Definitions 61.2.3 Task Models 71.2.4 Processor Models 81.2.5 Scheduling Models 91.2.6 Problem Decomposition 91.2.7 Types of Algorithms 101.3 Problem Analysis 101.3.1 Schedule Length Minimization 101.3.1.1 Uniprocessor computers 101.3.1.2 Multiprocessor computers 111.3.2 Energy Consumption Minimization 121.3.2.1 Uniprocessor computers 121.3.2.2 Multiprocessor computers 131.3.3 Strong NP-Hardness 141.3.4 Lower Bounds 141.3.5 Energy-Delay Trade-off 151.4 Pre-Power-Determination Algorithms 161.4.1 Overview 161.4.2 Performance Measures 171.4.3 Equal-Time Algorithms and Analysis 181.4.3.1 Schedule length minimization 181.4.3.2 Energy consumption minimization 191.4.4 Equal-Energy Algorithms and Analysis 191.4.4.1 Schedule length minimization 191.4.4.2 Energy consumption minimization 211.4.5 Equal-Speed Algorithms and Analysis 221.4.5.1 Schedule length minimization 221.4.5.2 Energy consumption minimization 231.4.6 Numerical Data 241.4.7 Simulation Results 251.5 Post-Power-Determination Algorithms 281.5.1 Overview 281.5.2 Analysis of List Scheduling Algorithms 291.5.2.1 Analysis of algorithm LS 291.5.2.2 Analysis of algorithm LRF 301.5.3 Application to Schedule Length Minimization 301.5.4 Application to Energy Consumption Minimization 311.5.5 Numerical Data 321.5.6 Simulation Results 321.6 Summary and Further Research 33References 342 POWER-AWARE HIGH PERFORMANCE COMPUTING 39Rong Ge and Kirk W. Cameron2.1 Introduction 392.2 Background 412.2.1 Current Hardware Technology and Power Consumption 412.2.1.1 Processor power 412.2.1.2 Memory subsystem power 422.2.2 Performance 432.2.3 Energy Efficiency 442.3 Related Work 452.3.1 Power Profiling 452.3.1.1 Simulator-based power estimation 452.3.1.2 Direct measurements 462.3.1.3 Event-based estimation 462.3.2 Performance Scalability on Power-Aware Systems 462.3.3 Adaptive Power Allocation for Energy-Efficient Computing 472.4 PowerPack: Fine-Grain Energy Profiling of HPC Applications 482.4.1 Design and Implementation of PowerPack 482.4.1.1 Overview 482.4.1.2 Fine-grain systematic power measurement 502.4.1.3 Automatic power profiling and code synchronization 512.4.2 Power Profiles of HPC Applications and Systems 532.4.2.1 Power distribution over components 532.4.2.2 Power dynamics of applications 542.4.2.3 Power bounds on HPC systems 552.4.2.4 Power versus dynamic voltage and frequency scaling 572.5 Power-Aware Speedup Model 592.5.1 Power-Aware Speedup 592.5.1.1 Sequential execution time for a single workload T1(w, f ) 602.5.1.2 Sequential execution time for an ON-chip/OFF-chip workload 602.5.1.3 Parallel execution time on N processors for an ON-/OFF-chip workload with DOP = i 612.5.1.4 Power-aware speedup for DOP and ON-/OFF-chip workloads 622.5.2 Model Parametrization and Validation 632.5.2.1 Coarse-grain parametrization and validation 642.5.2.2 Fine-grain parametrization and validation 662.6 Model Usages 692.6.1 Identification of Optimal System Configurations 702.6.2 PAS-Directed Energy-Driven Runtime Frequency Scaling 712.7 Conclusion 73References 753 ENERGY EFFICIENCY IN HPC SYSTEMS 81Ivan Rodero and Manish Parashar3.1 Introduction 813.2 Background and Related Work 833.2.1 CPU Power Management 833.2.1.1 OS-level CPU power management 833.2.1.2 Workload-level CPU power management 843.2.1.3 Cluster-level CPU power management 843.2.2 Component-Based Power Management 853.2.2.1 Memory subsystem 853.2.2.2 Storage subsystem 863.2.3 Thermal-Conscious Power Management 873.2.4 Power Management in Virtualized Datacenters 873.3 Proactive, Component-Based Power Management 883.3.1 Job Allocation Policies 883.3.2 Workload Profiling 903.4 Quantifying Energy Saving Possibilities 913.4.1 Methodology 923.4.2 Component-Level Power Requirements 923.4.3 Energy Savings 943.5 Evaluation of the Proposed Strategies 953.5.1 Methodology 963.5.2 Workloads 963.5.3 Metrics 973.6 Results 973.7 Concluding Remarks 1023.8 Summary 103References 1044 A STOCHASTIC FRAMEWORK FOR HIERARCHICAL SYSTEM-LEVEL POWER MANAGEMENT 109Peng Rong and Massoud Pedram4.1 Introduction 1094.2 Related Work 1114.3 A Hierarchical DPM Architecture 1134.4 Modeling 1144.4.1 Model of the Application Pool 1144.4.2 Model of the Service Flow Control 1184.4.3 Model of the Simulated Service Provider 1194.4.4 Modeling Dependencies between SPs 1204.5 Policy Optimization 1224.5.1 Mathematical Formulation 1224.5.2 Optimal Time-Out Policy for Local Power Manager 1234.6 Experimental Results 1254.7 Conclusion 130References 1305 ENERGY-EFFICIENT RESERVATION INFRASTRUCTURE FOR GRIDS, CLOUDS, AND NETWORKS 133Anne-Ce´ cile Orgerie and Laurent Lefe` vre5.1 Introduction 1335.2 Related Works 1345.2.1 Server and Data Center Power Management 1355.2.2 Node Optimizations 1355.2.3 Virtualization to Improve Energy Efficiency 1365.2.4 Energy Awareness in Wired Networking Equipment 1365.2.5 Synthesis 1375.3 ERIDIS: Energy-Efficient Reservation Infrastructure for Large-Scale Distributed Systems 1385.3.1 ERIDIS Architecture 1385.3.2 Management of the Resource Reservations 1415.3.3 Resource Management and On/Off Algorithms 1455.3.4 Energy-Consumption Estimates 1465.3.5 Prediction Algorithms 1465.4 EARI: Energy-Aware Reservation Infrastructure for Data Centers and Grids 1475.4.1 EARI’s Architecture 1475.4.2 Validation of EARI on Experimental Grid Traces 1475.5 GOC: Green Open Cloud 1495.5.1 GOC’s Resource Manager Architecture 1505.5.2 Validation of the GOC Framework 1525.6 HERMES: High Level Energy-Aware Model for Bandwidth Reservation in End-To-End Networks 1525.6.1 HERMES’ Architecture 1545.6.2 The Reservation Process of HERMES 1555.6.3 Discussion 1575.7 Summary 158References 1586 ENERGY-EFFICIENT JOB PLACEMENT ON CLUSTERS, GRIDS, AND CLOUDS 163Damien Borgetto, Henri Casanova, Georges Da Costa, and Jean-Marc Pierson6.1 Problem and Motivation 1636.1.1 Context 1636.1.2 Chapter Roadmap 1646.2 Energy-Aware Infrastructures 1646.2.1 Buildings 1656.2.2 Context-Aware Buildings 1656.2.3 Cooling 1666.3 Current Resource Management Practices 1676.3.1 Widely Used Resource Management Systems 1676.3.2 Job Requirement Description 1696.4 Scientific and Technical Challenges 1706.4.1 Theoretical Difficulties 1706.4.2 Technical Difficulties 1706.4.3 Controlling and Tuning Jobs 1716.5 Energy-Aware Job Placement Algorithms 1726.5.1 State of the Art 1726.5.2 Detailing One Approach 1746.6 Discussion 1806.6.1 Open Issues and Opportunities 1806.6.2 Obstacles for Adoption in Production 1826.7 Conclusion 183References 1847 COMPARISON AND ANALYSIS OF GREEDY ENERGY-EFFICIENT SCHEDULING ALGORITHMS FOR COMPUTATIONAL GRIDS 189Peder Lindberg, James Leingang, Daniel Lysaker, Kashif Bilal, Samee Ullah Khan, Pascal Bouvry, Nasir Ghani, Nasro Min-Allah, and Juan Li7.1 Introduction 1897.2 Problem Formulation 1917.2.1 The System Model 1917.2.1.1 PEs 1917.2.1.2 DVS 1917.2.1.3 Tasks 1927.2.1.4 Preliminaries 1927.2.2 Formulating the Energy-Makespan Minimization Problem 1927.3 Proposed Algorithms 1937.3.1 Greedy Heuristics 1947.3.1.1 Greedy heuristic scheduling algorithm 1967.3.1.2 Greedy-min 1977.3.1.3 Greedy-deadline 1987.3.1.4 Greedy-max 1987.3.1.5 MaxMin 1997.3.1.6 ObFun 1997.3.1.7 MinMin StdDev 2027.3.1.8 MinMax StdDev 2027.4 Simulations, Results, and Discussion 2037.4.1 Workload 2037.4.2 Comparative Results 2047.4.2.1 Small-size problems 2047.4.2.2 Large-size problems 2067.5 Related Works 2117.6 Conclusion 211References 2128 TOWARD ENERGY-AWARE SCHEDULING USING MACHINE LEARNING 215Josep LL. Berral, In˜ igo Goiri, Ramon Nou, Ferran Julia` , Josep O. Fito´ , Jordi Guitart, Ricard Gavalda´ , and Jordi Torres8.1 Introduction 2158.1.1 Energetic Impact of the Cloud 2168.1.2 An Intelligent Way to Manage Data Centers 2168.1.3 Current Autonomic Computing Techniques 2178.1.4 Power-Aware Autonomic Computing 2178.1.5 State of the Art and Case Study 2188.2 Intelligent Self-Management 2188.2.1 Classical AI Approaches 2198.2.1.1 Heuristic algorithms 2198.2.1.2 AI planning 2198.2.1.3 Semantic techniques 2198.2.1.4 Expert systems and genetic algorithms 2208.2.2 Machine Learning Approaches 2208.2.2.1 Instance-based learning 2218.2.2.2 Reinforcement learning 2228.2.2.3 Feature and example selection 2258.3 Introducing Power-Aware Approaches 2258.3.1 Use of Virtualization 2268.3.2 Turning On and Off Machines 2288.3.3 Dynamic Voltage and Frequency Scaling 2298.3.4 Hybrid Nodes and Data Centers 2308.4 Experiences of Applying ML on Power-Aware Self-Management 2308.4.1 Case Study Approach 2318.4.2 Scheduling and Power Trade-Off 2318.4.3 Experimenting with Power-Aware Techniques 2338.4.4 Applying Machine Learning 2368.4.5 Conclusions from the Experiments 2388.5 Conclusions on Intelligent Power-Aware Self-Management 238References 2409 ENERGY EFFICIENCY METRICS FOR DATA CENTERS 245Javid Taheri and Albert Y. Zomaya9.1 Introduction 2459.1.1 Background 2459.1.2 Data Center Energy Use 2469.1.3 Data Center Characteristics 2469.1.3.1 Electric power 2479.1.3.2 Heat removal 2499.1.4 Energy Efficiency 2509.2 Fundamentals of Metrics 2509.2.1 Demand and Constraints on Data Center Operators 2509.2.2 Metrics 2519.2.2.1 Criteria for good metrics 2519.2.2.2 Methodology 2529.2.2.3 Stability of metrics 2529.3 Data Center Energy Efficiency 2529.3.1 Holistic IT Efficiency Metrics 2529.3.1.1 Fixed versus proportional overheads 2549.3.1.2 Power versus energy 2549.3.1.3 Performance versus productivity 2559.3.2 Code of Conduct 2569.3.2.1 Environmental statement 2569.3.2.2 Problem statement 2569.3.2.3 Scope of the CoC 2579.3.2.4 Aims and objectives of CoC 2589.3.3 Power Use in Data Centers 2599.3.3.1 Data center IT power to utility power relationship 2599.3.3.2 Chiller efficiency and external temperature 2609.4 Available Metrics 2609.4.1 The Green Grid 2619.4.1.1 Power usage effectiveness (PUE) 2619.4.1.2 Data center efficiency (DCE) 2629.4.1.3 Data center infrastructure efficiency (DCiE) 2629.4.1.4 Data center productivity (DCP) 2639.4.2 McKinsey 2639.4.3 Uptime Institute 2649.4.3.1 Site infrastructure power overhead multiplier (SI-POM) 2659.4.3.2 IT hardware power overhead multiplier (H-POM) 2669.4.3.3 DC hardware compute load per unit of computing work done 2669.4.3.4 Deployed hardware utilization ratio (DH-UR) 2669.4.3.5 Deployed hardware utilization efficiency (DH-UE) 2679.5 Harmonizing Global Metrics for Data Center Energy Efficiency 267References 26810 AUTONOMIC GREEN COMPUTING IN LARGE-SCALE DATA CENTERS 271Haoting Luo, Bithika Khargharia, Salim Hariri, and Youssif Al-Nashif 10.1 Introduction 27110.2 Related Technologies and Techniques 27210.2.1 Power Optimization Techniques in Data Centers 27210.2.2 Design Model 27310.2.3 Networks 27410.2.4 Data Center Power Distribution 27510.2.5 Data Center Power-Efficient Metrics 27610.2.6 Modeling Prototype and Testbed 27710.2.7 Green Computing 27810.2.8 Energy Proportional Computing 28010.2.9 Hardware Virtualization Technology 28110.2.10 Autonomic Computing 28210.3 Autonomic Green Computing: A Case Study 28310.3.1 Autonomic Management Platform 28510.3.1.1 Platform architecture 28510.3.1.2 DEVS-based modeling and simulation platform 28510.3.1.3 Workload generator 28710.3.2 Model Parameter Evaluation 28810.3.2.1 State transitioning overhead 28810.3.2.2 VM template evaluation 28910.3.2.3 Scalability analysis 29110.3.3 Autonomic Power Efficiency Management Algorithm (Performance Per Watt) 29110.3.4 Simulation Results and Evaluation 29310.3.4.1 Analysis of energy and performance trade-offs 29610.4 Conclusion and Future Directions 297References 29811 ENERGY AND THERMAL AWARE SCHEDULING IN DATA CENTERS 301Gaurav Dhiman, Raid Ayoub, and Tajana S. Rosing11.1 Introduction 30111.2 Related Work 30211.3 Intermachine Scheduling 30511.3.1 Performance and Power Profile of VMs 30511.3.2 Architecture 30911.3.2.1 vgnode 30911.3.2.2 vgxen 31011.3.2.3 vgdom 31211.3.2.4 vgserv 31211.4 Intramachine Scheduling 31511.4.1 Air-Forced Thermal Modeling and Cost 31611.4.2 Cooling Aware Dynamic Workload Scheduling 31711.4.3 Scheduling Mechanism 31811.4.4 Cooling Costs Predictor 31911.5 Evaluation 32111.5.1 Intermachine Scheduler (vGreen) 32111.5.2 Heterogeneous Workloads 32311.5.2.1 Comparison with DVFS policies 32511.5.2.2 Homogeneous workloads 32811.5.3 Intramachine Scheduler (Cool and Save) 32811.5.3.1 Results 33111.5.3.2 Overhead of CAS 33311.6 Conclusion 333References 33412 QOS-AWARE POWER MANAGEMENT IN DATA CENTERS 339Jiayu Gong and Cheng-Zhong Xu12.1 Introduction 33912.2 Problem Classification 34012.2.1 Objective and Constraint 34012.2.2 Scope and Time Granularities 34012.2.3 Methodology 34112.2.4 Power Management Mechanism 34212.3 Energy Efficiency 34412.3.1 Energy-Efficiency Metrics 34412.3.2 Improving Energy Efficiency 34612.3.2.1 Energy minimization with performance guarantee 34612.3.2.2 Performance maximization under power budget 34812.3.2.3 Trade-off between power and performance 34812.3.3 Energy-Proportional Computing 35012.4 Power Capping 35112.5 Conclusion 353References 35613 ENERGY-EFFICIENT STORAGE SYSTEMS FOR DATA CENTERS 361Sudhanva Gurumurthi and Anand Sivasubramaniam13.1 Introduction 36113.2 Disk Drive Operation and Disk Power 36213.2.1 An Overview of Disk Drives 36213.2.2 Sources of Disk Power Consumption 36313.2.3 Disk Activity and Power Consumption 36513.3 Disk and Storage Power Reduction Techniques 36613.3.1 Exploiting the STANDBY State 36813.3.2 Reducing Seek Activity 36913.3.3 Achieving Energy Proportionality 36913.3.3.1 Hardware approaches 36913.3.3.2 Software approaches 37013.4 Using Nonvolatile Memory and Solid-State Disks 37113.5 Conclusions 372References 37314 AUTONOMIC ENERGY/PERFORMANCE OPTIMIZATIONS FOR MEMORY IN SERVERS 377Bithika Khargharia and Mazin Yousif14.1 Introduction 37814.2 Classifications of Dynamic Power Management Techniques 38014.2.1 Heuristic and Predictive Techniques 38014.2.2 QoS and Energy Trade-Offs 38114.3 Applications of Dynamic Power Management (DPM) 38214.3.1 Power Management of System Components in Isolation 38214.3.2 Joint Power Management of System Components 38314.3.3 Holistic System-Level Power Management 38314.4 Autonomic Power and Performance Optimization of Memory Subsystems in Server Platforms 38414.4.1 Adaptive Memory Interleaving Technique for Power and Performance Management 38414.4.1.1 Formulating the optimization problem 38614.4.1.2 Memory appflow 38914.4.2 Industry Techniques 38914.4.2.1 Enhancements in memory hardware design 39014.4.2.2 Adding more operating states 39014.4.2.3 Faster transition to and from low power states 39014.4.2.4 Memory consolidation 39014.5 Conclusion 391References 39115 ROD: A PRACTICAL APPROACH TO IMPROVING RELIABILITY OF ENERGY-EFFICIENT PARALLEL DISK SYSTEMS 395Shu Yin, Xiaojun Ruan, Adam Manzanares, and Xiao Qin15.1 Introduction 39515.2 Modeling Reliability of Energy-Efficient Parallel Disks 39615.2.1 The MINT Model 39615.2.1.1 Disk utilization 39815.2.1.2 Temperature 39815.2.1.3 Power-state transition frequency 39915.2.1.4 Single disk reliability model 39915.2.2 MAID, Massive Arrays of Idle Disks 40015.3 Improving Reliability of MAID via Disk Swapping 40115.3.1 Improving Reliability of Cache Disks in MAID 40115.3.2 Swapping Disks Multiple Times 40415.4 Experimental Results and Evaluation 40515.4.1 Experimental Setup 40515.4.2 Disk Utilization 40615.4.3 The Single Disk Swapping Strategy 40615.4.4 The Multiple Disk Swapping Strategy 40915.5 Related Work 41115.6 Conclusions 412References 41316 EMBRACING THE MEMORY AND I/O WALLS FOR ENERGY-EFFICIENT SCIENTIFIC COMPUTING 417Chung-Hsing Hsu and Wu-Chun Feng16.1 Introduction 41716.2 Background and Related Work 42016.2.1 DVFS-Enabled Processors 42016.2.2 DVFS Scheduling Algorithms 42116.2.3 Memory-Aware, Interval-Based Algorithms 42216.3 β-Adaptation: A New DVFS Algorithm 42316.3.1 The Compute-Boundedness Metric, β 42316.3.2 The Frequency Calculating Formula, f ∗ 42416.3.3 The Online β Estimation 42516.3.4 Putting It All Together 42716.4 Algorithm Effectiveness 42916.4.1 A Comparison to Other DVFS Algorithms 42916.4.2 Frequency Emulation 43216.4.3 The Minimum Dependence to the PMU 43616.5 Conclusions and Future Work 438References 43917 MULTIPLE FREQUENCY SELECTION IN DVFS-ENABLED PROCESSORS TO MINIMIZE ENERGY CONSUMPTION 443Nikzad Babaii Rizvandi, Albert Y. Zomaya, Young Choon Lee, Ali Javadzadeh Boloori, and Javid Taheri17.1 Introduction 44317.2 Energy Efficiency in HPC Systems 44417.3 Exploitation of Dynamic Voltage–Frequency Scaling 44617.3.1 Independent Slack Reclamation 44617.3.2 Integrated Schedule Generation 44717.4 Preliminaries 44817.4.1 System and Application Models 44817.4.2 Energy Model 44817.5 Energy-Aware Scheduling via DVFS 45017.5.1 Optimum Continuous Frequency 45017.5.2 Reference Dynamic Voltage–Frequency Scaling (RDVFS) 45117.5.3 Maximum-Minimum-Frequency for Dynamic Voltage–Frequency Scaling (MMF-DVFS) 45217.5.4 Multiple Frequency Selection for Dynamic Voltage–Frequency Scaling (MFS-DVFS) 45317.5.4.1 Task eligibility 45417.6 Experimental Results 45617.6.1 Simulation Settings 45617.6.2 Results 45817.7 Conclusion 461References 46118 THE PARAMOUNTCY OF RECONFIGURABLE COMPUTING 465Reiner Hartenstein18.1 Introduction 46518.2 Why Computers are Important 46618.2.1 Computing for a Sustainable Environment 47018.3 Performance Progress Stalled 47218.3.1 Unaffordable Energy Consumption of Computing 47318.3.2 Crashing into the Programming Wall 47518.4 The Tail is Wagging the Dog (Accelerators) 48818.4.1 Hardwired Accelerators 48918.4.2 Programmable Accelerators 49018.5 Reconfigurable Computing 49418.5.1 Speedup Factors by FPGAs 49818.5.2 The Reconfigurable Computing Paradox 50118.5.3 Saving Energy by Reconfigurable Computing 50518.5.3.1 Traditional green computing 50618.5.3.2 The role of graphics processors 50718.5.3.3 Wintel versus ARM 50818.5.4 Reconfigurable Computing is the Silver Bullet 51118.5.4.1 A new world model of computing 51118.5.5 The Twin-Paradigm Approach to Tear Down the Wall 51418.5.6 A Mass Movement Needed as Soon as Possible 51718.5.6.1 Legacy software from the mainframe age 51818.5.7 How to Reinvent Computing 519 18.6 Conclusions 526References 52919 WORKLOAD CLUSTERING FOR INCREASING ENERGY SAVINGS ON EMBEDDED MPSOCS 549Ozcan Ozturk, Mahmut Kandemir, and Sri Hari Krishna Narayanan19.1 Introduction 54919.2 Embedded MPSoC Architecture, Execution Model, and Related Work 55019.3 Our Approach 55119.3.1 Overview 55119.3.2 Technical Details and Problem Formulation 55319.3.2.1 System and job model 55319.3.2.2 Mathematical programing model 55419.3.2.3 Example 55719.4 Experimental Evaluation 56019.5 Conclusions 564References 56520 ENERGY-EFFICIENT INTERNET INFRASTRUCTURE 567Weirong Jiang and Viktor K. Prasanna20.1 Introduction 56720.1.1 Performance Challenges 56820.1.2 Existing Packet Forwarding Approaches 57020.1.2.1 Software approaches 57020.1.2.2 Hardware approaches 57120.2 SRAM-Based Pipelined IP Lookup Architectures: Alternative to TCAMs 57120.3 Data Structure Optimization for Power Efficiency 57320.3.1 Problem Formulation 57420.3.1.1 Non-pipelined and pipelined engines 57420.3.1.2 Power function of SRAM 57520.3.2 Special Case: Uniform Stride 57620.3.3 Dynamic Programming 57620.3.4 Performance Evaluation 57720.3.4.1 Results for non-pipelined architecture 57820.3.4.2 Results for pipelined architecture 57820.4 Architectural Optimization to Reduce Dynamic Power Dissipation 58020.4.1 Analysis and Motivation 58120.4.1.1 Traffic locality 58220.4.1.2 Traffic rate variation 58220.4.1.3 Access frequency on different stages 58320.4.2 Architecture-Specific Techniques 58320.4.2.1 Inherent caching 58420.4.2.2 Local clocking 58420.4.2.3 Fine-grained memory enabling 58520.4.3 Performance Evaluation 58520.5 Related Work 58820.6 Summary 589References 58921 DEMAND RESPONSE IN THE SMART GRID: A DISTRIBUTED COMPUTING PERSPECTIVE 593Chen Wang and Martin De Groot21.1 Introduction 59321.2 Demand Response 59521.2.1 Existing Demand Response Programs 59521.2.2 Demand Response Supported by the Smart Grid 59721.3 Demand Response as a Distributed System 60021.3.1 An Overlay Network for Demand Response 60021.3.2 Event Driven Demand Response 60221.3.3 Cost Driven Demand Response 60421.3.4 A Decentralized Demand Response Framework 60921.3.5 Accountability of Coordination Decision Making 61021.4 Summary 611References 61122 RESOURCE MANAGEMENT FOR DISTRIBUTED MOBILE COMPUTING 615Jong-Kook Kim22.1 Introduction 61522.2 Single-Hop Energy-Constrained Environment 61722.2.1 System Model 61722.2.2 Related Work 62022.2.3 Heuristic Descriptions 62122.2.3.1 Mapping event 62122.2.3.2 Scheduling communications 62122.2.3.3 Opportunistic load balancing and minimum energy greedy heuristics 62222.2.3.4 ME-MC heuristic 62222.2.3.5 ME-ME heuristic 62422.2.3.6 CRME heuristic 62522.2.3.7 Originator and random 62622.2.3.8 Upper bound 62622.2.4 Simulation Model 62822.2.5 Results 63022.2.6 Summary 63422.3 Multihop Distributed Mobile Computing Environment 63522.3.1 The Multihop System Model 63522.3.2 Energy-Aware Routing Protocol 63622.3.2.1 Overview 63622.3.2.2 DSDV 63722.3.2.3 DSDV remaining energy 63722.3.2.4 DSDV-energy consumption per remaining energy 63722.3.3 Heuristic Description 63822.3.3.1 Random 63822.3.3.2 Estimated minimum total energy (EMTE) 63822.3.3.3 K-percent-speed (KPS) and K-percent-energy (KPE) 63922.3.3.4 Energy ratio and distance (ERD) 63922.3.3.5 ETC and distance (ETCD) 64022.3.3.6 Minimum execution time (MET) 64022.3.3.7 Minimum completion time (MCT) and minimum completion time with DVS (MCT-DVS) 64022.3.3.8 Switching algorithm (SA) 64022.3.4 Simulation Model 64122.3.5 Results 64322.3.5.1 Distributed resource management 64322.3.5.2 Energy-aware protocol 64422.3.6 Summary 64422.4 Future Work 647References 64723 AN ENERGY-AWARE FRAMEWORK FOR MOBILE DATA MINING 653Carmela Comito, Domenico Talia, and Paolo Trunfio23.1 Introduction 65323.2 System Architecture 65423.3 Mobile Device Components 65723.4 Energy Model 65923.5 Clustering Scheme 66423.5.1 Clustering the M2M Architecture 66623.6 Conclusion 670References 67024 ENERGY AWARENESS AND EFFICIENCY IN WIRELESS SENSOR NETWORKS: FROM PHYSICAL DEVICES TO THE COMMUNICATION LINK 673Fla´ via C. Delicato and Paulo F. Pires24.1 Introduction 67324.2 WSN and Power Dissipation Models 67624.2.1 Network and Node Architecture 67624.2.2 Sources of Power Dissipation in WSNs 67924.3 Strategies for Energy Optimization 68324.3.1 Intranode Level 68424.3.1.1 Duty cycling 68524.3.1.2 Adaptive sensing 69124.3.1.3 Dynamic voltage scale (DVS) 69324.3.1.4 OS task scheduling 69424.3.2 Internode Level 69524.3.2.1 Transmission power control 69524.3.2.2 Dynamic modulation scaling 69624.3.2.3 Link layer optimizations 69824.4 Final Remarks 701References 70225 NETWORK-WIDE STRATEGIES FOR ENERGY EFFICIENCY IN WIRELESS SENSOR NETWORKS 709Fla´ via C. Delicato and Paulo F. Pires25.1 Introduction 70925.2 Data Link Layer 71125.2.1 Topology Control Protocols 71225.2.2 Energy-Efficient MAC Protocols 71425.2.2.1 Scheduled MAC protocols in WSNs 71625.2.2.2 Contention-based MAC protocols 71725.3 Network Layer 71925.3.1 Flat and Hierarchical Protocols 72225.4 Transport Layer 72525.5 Application Layer 72925.5.1 Task Scheduling 72925.5.2 Data Aggregation and Data Fusion in WSNs 73325.5.2.1 Approaches of data fusion for energy efficiency 73525.5.2.2 Data aggregation strategies 73625.6 Final Remarks 740References 74126 ENERGY MANAGEMENT IN HETEROGENEOUS WIRELESS HEALTH CARE NETWORKS 751Nima Nikzad, Priti Aghera, Piero Zappi, and Tajana S. Rosing26.1 Introduction 75126.2 System Model 75326.2.1 Health Monitoring Task Model 75326.3 Collaborative Distributed Environmental Sensing 75526.3.1 Node Neighborhood and Localization Rate 75726.3.2 Energy Ratio and Sensing Rate 75826.3.3 Duty Cycling and Prediction 75926.4 Task Assignment in a Body Area Network 76026.4.1 Optimal Task Assignment 76026.4.2 Dynamic Task Assignment 76226.4.2.1 DynAGreen algorithm 76326.4.2.2 DynAGreenLife algorithm 76826.5 Results 77126.5.1 Collaborative Sensing 77126.5.1.1 Results 77226.5.2 Dynamic Task Assignment 77626.5.2.1 Performance in static conditions 77726.5.2.2 Dynamic adaptability 78026.6 Conclusion 784References 785INDEX 787