The 5G technology has been commercialized worldwide and is expected to provide superior performance with enhanced mobile broadband, ultra-low latency transmission, and massive IoT connections. Meanwhile, the edge computing paradigm gets popular to provide distributed computing and storage resources in proximity to the users. As edge services and applications prosper, 5G and edge computing will be tightly coupled and continuously promote each other forward.
Embracing this trend, however, mobile users, infrastructure providers, and service providers are all faced with the energy dilemma. On the user side, battery-powered mobile devices are much constrained by battery life, whereas mobile platforms and apps nowadays are usually power-hungry. At the infrastructure and service provider side, the energy cost of edge facilities accounts for a large proportion of operating expenses and has become a huge burden.
This book provides a collection of most recent attempts to tackle the energy issues in mobile edge computing from new and promising perspectives. For example, the book investigates the pervasive low-battery anxiety among modern mobile users and quantifies the anxiety degree and likely behavior concerning the battery status. Based on the quantified model, a low-power video streaming solution is developed accordingly to save mobile devices' energy and alleviate users' low-battery anxiety. In addition to energy management for mobile users, the book also looks into potential opportunities to energy cost saving and carbon emission reduction at edge facilities, particularly the 5G base stations and geo-distributed edge datacenters.
Author(s): Guoming Tang, Deke Guo, Kui Wu
Series: SpringerBriefs in Computer Science
Publisher: Springer
Year: 2022
Language: English
Pages: 118
City: Singapore
Preface
Contents
1 Introduction
1.1 When 5G Meets Edge Computing
1.2 The Energy Dilemma
1.3 Key Problems and Contributions
1.4 Content Organization
2 Investigating Low-Battery Anxiety of Mobile Users
2.1 Introduction
2.2 Related Work
2.3 A Survey Over 2000+ Mobile Users
2.4 Quantification of Low-Battery Anxiety
2.4.1 Extraction of LBA Curve
2.4.2 Observations and Analysis
2.4.2.1 Overall LBA Curve
2.4.2.2 LBA Curves for Different Age Groups
2.4.2.3 Role of Gender in Battery Charging
2.4.3 Lessons Learnt from LBA Quantification
2.5 Impacts of LBA on Video Watching
2.5.1 Extraction of Video Abandoning Likelihood Curve
2.5.2 Observations and Analysis
2.5.2.1 Overall Video Abandoning Likelihood
2.5.2.2 Video Abandoning Likelihood for Different Age Groups
2.5.2.3 Correlation Between Battery Charging and Video Abandoning
2.5.3 Advice for Video Streaming Services
2.6 Ethics
2.7 Conclusion
3 User Energy and LBA Aware Mobile Video Streaming
3.1 Introduction
3.2 Background and Related Work
3.2.1 Background of Low-Battery Anxiety
3.2.2 Background of Display Power Saving
3.2.2.1 Power Saving for LCD
3.2.2.2 Power Saving for OLED
3.2.3 Work Related to This Work
3.2.3.1 Proxy-Based Energy Saving
3.2.3.2 Energy-Performance Tradeoff
3.3 LBA Survey and Modelling
3.3.1 Data Collection
3.3.2 LBA Curve Extraction
3.3.3 Insights on LBA Alleviation
3.4 LPVS: Low-Power Video Streaming
3.4.1 Scenario Overview
3.4.2 Models for Power Consumption in Video Streaming
3.4.3 Models for Energy Status and Low-Battery Anxiety
3.4.4 Video Streaming Capacity at the Edge
3.4.5 Joint Optimization for Energy Saving and Anxiety Reduction
3.5 Solution Methodology
3.5.1 The Difficulties
3.5.2 Information Compacting
3.5.3 A Two-Phase Heuristic for Joint Optimization
3.5.4 Determine γn with Bayesian Inference
3.6 LBA Model Updating
3.6.1 Analysis of LBA Heterogeneity
3.6.2 Local LBA Model Updating
3.7 Implementations
3.7.1 Real-World Video Streaming Traces
3.7.2 LPVS Emulation and Setups
3.8 Performance Evaluations
3.8.1 LPVS with Sufficient Edge Resource
3.8.1.1 Energy Saving of Mobile Devices
3.8.1.2 Anxiety Reduction of Mobile Users
3.8.2 LPVS with Limited Edge Resource
3.8.2.1 Energy Saving of Mobile Devices
3.8.2.2 Anxiety Reduction of Mobile Users
3.8.3 Impact of LPVS on Low-Battery Users
3.8.4 LPVS with Updated LBA Models
3.8.5 Overhead of LPVS and Impact on Other QoE Metrics
3.9 Conclusion
4 Optimal Backup Power Allocation for 5G Base Stations
4.1 Introduction
4.1.1 Spatial Dimension
4.1.2 Temporal Dimension
4.2 Related Work
4.3 BS Power Measurements and Observations
4.3.1 Power Consumption of 4G and 5G BSs
4.3.2 Power Consumption of 5G BS Major Components
4.3.3 Multiplexing Gain with Backup Power Sharing
4.4 System Model
4.4.1 Scenario Overview
4.4.1.1 Base Stations
4.4.1.2 Backup Batteries
4.4.1.3 Battery Capacity and Deployment
4.4.2 Traffic Load and Power Demand
4.5 Optimal Backup Power Allocation
4.5.1 Analysis of Power Outages and Network Failure
4.5.1.1 Synchronous Outage
4.5.1.2 Asynchronous Outage
4.5.2 Condition of Network Reliability
4.5.3 Backup Power Deployment Constraints
4.5.4 Backup Power Allocation Optimization
4.6 Experimental Evaluations
4.6.1 Experiment Setup
4.6.1.1 Scenario
4.6.1.2 BS Power Demands
4.6.1.3 Benchmarks
4.6.2 Results and Analysis
4.6.2.1 Cost Saving vs. Number of PoPs
4.6.2.2 Cost Saving vs. C and L
4.6.2.3 Case Study
4.7 Conclusion
5 Reusing Backup Batteries for Power Demand Reshaping in 5G
5.1 Introduction
5.2 System Models
5.2.1 Scenario Overview
5.2.2 BS Power Supply and Demand
5.2.3 Battery Specification
5.3 Power Demand Reshaping via BESS Scheduling
5.3.1 Energy Cost with BESS
5.3.2 Battery Degradation Cost
5.3.3 Optimal BESS Operation Scheduling
5.3.3.1 Discharge/Charge Rate Constraint
5.3.3.2 SoC Constraint
5.3.3.3 Battery State Updating
5.3.4 Problem Analysis
5.3.4.1 High Computational Complexity
5.3.4.2 Dynamic Power Demand
5.4 A DRL-Based Approach to Distributed BESS Scheduling
5.4.1 DRL Based BESS Scheduling: Components and Concepts
5.4.2 Reward Function Design
5.4.3 Learning Process Design
5.5 Experimental Evaluations
5.5.1 Experiment Setup
5.5.1.1 BS and Traffic Demand Data
5.5.1.2 Parameter Settings
5.5.1.3 Scenario Settings
5.5.2 General Performance at Cost Reduction with BESS
5.5.2.1 All w/BESS
5.5.2.2 Part w/BESS
5.5.3 Case Studies of DRL-Based BESS Scheduling
5.5.3.1 From an Overall View
5.5.3.2 From Local Views
5.5.4 ROIs of Different BESS Deployments
5.6 Related Work
5.6.1 General System Peak Power Shaving with BESS
5.6.2 DC Peak Power Shaving with Centralized BESS
5.6.3 DC Peak Power Shaving with Distributed BESS
5.7 Conclusion
6 Software-Defined Power Supply to Geo-Distributed Edge DCs
6.1 Introduction
6.2 Architecture of Software-Defined Power Supply (SDPS)
6.2.1 Motivation and Design Rationales
6.2.2 Architecture Design
6.3 Two-Phase Optimization in Software-Defined Power Supply
6.3.1 System Model
6.3.2 Phase-I: Constructing Green Cells
6.3.3 Phase-II: BESS Discharging/Charging Operations
6.4 Experimental Evaluations
6.4.1 Experiments Setup
6.4.2 Performance Comparison
6.5 Conclusion
7 Conclusions and Future Work
7.1 Conclusions
7.2 Future Work
A Questionnaire of LBA Survey and Collected Answers
Bibliography
