Content Distribution for Mobile Internet: A Cloud-based Approach

Description

Content distribution, i.e., distributing digital content from one node to another node or multiple nodes, is the most fundamental function of the Internet. Since Amazon’s launch of EC2 in 2006 and Apple’s release of the iPhone in 2007, Internet content distribution has shown a strong trend toward polarization. On the one hand, considerable investments have been made in creating heavyweight, integrated data centers (“heavy-cloud”) all over the world, in order to achieve economies of scale and high flexibility/efficiency of content distribution. On the other hand, end-user devices (“light-end”) have become increasingly lightweight, mobile and heterogeneous, creating new demands concerning traffic usage, energy consumption, bandwidth, latency, reliability, and/or the security of content distribution. Based on comprehensive real-world measurements at scale, we observe that existing content distribution techniques often perform poorly under the abovementioned new circumstances. Motivated by the trend of “heavy-cloud vs. light-end,” this book is dedicated to uncovering the root causes of today’s mobile networking problems and designing innovative cloud-based solutions to practically address such problems. Our work has produced not only academic papers published in prestigious conference proceedings like SIGCOMM, NSDI, MobiCom and MobiSys, but also concrete effects on industrial systems such as Xiaomi Mobile, MIUI OS, Tencent App Store, Baidu PhoneGuard, and WiFi.com. A series of practical takeaways and easy-to-follow testimonials are provided to researchers and practitioners working in mobile networking and cloud computing. In addition, we have released as much code and data used in our research as possible to benefit the community. Preface Acknowledgments Contents About the Authors Acronyms Part I Get Started 1 Background and Overview 1.1 Internet Content Distribution 1.2 Cloud Computing and Mobile Internet 1.3 Frontier Techniques 1.4 Overview of the Book Structure References Part II Metric 1 Speed: Bandwidth Is the First 2 Fast and Light Bandwidth Testing for Mobile Internet 2.1 Introduction 2.2 Understanding State-of-the-Art BTSes 2.2.1 Methodology 2.2.2 Analyzing Deployed BTSes 2.2.3 Measurement Results 2.2.4 Case Studies 2.3 Design of FastBTS 2.3.1 Crucial Interval Sampling (CIS) 2.3.2 Elastic Bandwidth Probing (EBP) 2.3.3 Data-Driven Server Selection (DSS) 2.3.4 Adaptive Multi-Homing (AMH) 2.4 Implementation 2.5 Evaluation 2.5.1 Experiment Setup 2.5.2 End-to-End Performance 2.5.3 Individual Components 2.6 Concluding Remarks References 3 Offline Downloading via Cloud and/or Smart APs 3.1 Introduction 3.2 Related Work 3.3 System Overview 3.3.1 Overview of Xuanfeng 3.3.2 Overview of the Smart AP Systems 3.4 Workload Characteristics 3.5 Performance of the Cloud-Based System 3.5.1 Pre-downloading Performance 3.5.2 Fetching Performance 3.5.3 End-to-End Performance 3.6 Performance of the Smart APs 3.6.1 Methodology 3.6.2 Benchmark Results 3.7 The ODR Middleware 3.7.1 Design and Implementation 3.7.2 Performance Evaluation 3.8 Conclusion References Part III Metric 2 Cost: Never Waste on Traffic Usage 4 Cross-Application Cellular Traffic Optimization 4.1 Introduction 4.2 State-of-the-Art Systems 4.3 Measuring Cellular Traffic 4.3.1 Dataset Collection 4.3.2 Content Analysis 4.3.2.1 General Characteristics 4.3.2.2 Size and Quality of Content 4.3.2.3 Content Validation 4.3.2.4 Traffic Filtering 4.3.2.5 Caching Strategies 4.4 System Overview 4.5 Mechanisms 4.5.1 Image Compression 4.5.2 Content Validation 4.5.3 Traffic Filtering 4.5.4 Value-Based Web Caching (VBWC) 4.6 Evaluation 4.6.1 Data Collection and Methodology 4.6.2 Traffic Reduction 4.6.3 System Overhead 4.6.4 Latency Penalty 4.7 Conclusion References 5 Boosting Mobile Virtual Network Operator 5.1 Introduction 5.2 Background 5.3 Measurement Data Collection 5.4 Data Usage Characterization 5.5 Network Performance 5.6 Data Usage Prediction and Data Reselling Optimization 5.6.1 Data Usage Prediction and Modeling 5.6.2 Data Reselling Optimization 5.7 Customer Churn Profiling and Mitigation 5.8 Inaccurate Billing in MVNO 5.9 Related Work 5.10 Concluding Remarks References Part IV Metric 3 Reliability: Stay Connected All the Time 6 Enhancing Nationwide Cellular Reliability 6.1 Introduction 6.2 Study Methodology 6.2.1 Limitations of Vanilla Android 6.2.2 Continuous Monitoring Infrastructure 6.2.3 Large-Scale Deployment 6.3 Measurement Results 6.3.1 General Statistics 6.3.2 Android Phone Landscape 6.3.3 ISP and Base Station Landscape 6.4 Enhancements 6.4.1 Guidelines in Principle 6.4.2 Real-World Practices 6.4.3 Deployment and Evaluation 6.5 Conclusion References 7 Legal Internet Access Under Extreme Censorship 7.1 Introduction 7.2 Extreme Internet Censorship 7.3 The ScholarCloud System 7.4 Measurement Study 7.4.1 Common Practices 7.4.2 Methodology 7.4.3 Measurement Results 7.5 Related Work 7.6 Limitation and Discussion References Part V Metric 4 Security: Do Not Relax Our Vigilance 8 Combating Nationwide WiFi Security Threats 8.1 Introduction 8.2 Study Methodology 8.2.1 WiSC System Overview 8.2.2 LAN Attack Detection 8.2.3 WAN Attack Detection 8.2.3.1 Transport-Layer Attack Detection 8.2.3.2 Data-Driven Parameter Settings 8.2.3.3 Application-Layer Detection 8.2.4 Evaluation 8.2.5 Large-Scale Deployment and Field Survey 8.3 Measurement Results 8.3.1 Prevalence of WiFi Attacks 8.3.2 WiFi-Based Attack Techniques 8.3.3 Malicious Behaviors and Objectives 8.3.4 Fundamental Motives Behind the Attacks 8.4 Undermining the Attack Ecosystem 8.4.1 Uncovering the Underground Ecosystem 8.4.2 Interplay and Weakness 8.4.3 Real-World Active Defense Practices 8.5 Conclusion References 9 Understanding IoT Security with HoneyCloud 9.1 Introduction 9.2 Honeypot Deployment 9.2.1 Overview 9.2.2 Hardware IoT Honeypots 9.2.3 Software IoT Honeypots 9.2.3.1 High Fidelity Maintainer 9.2.3.2 Shell Interceptor and Inference Terminal 9.2.3.3 Access Controller 9.2.3.4 Reset Manager 9.3 Findings and Implications 9.3.1 General Characteristics and Statistics 9.3.2 Malware-Based Attacks 9.3.3 Fileless Attack Taxonomy 9.3.4 Key Insights for Fileless Attacks 9.3.5 New Security Challenges and Defense Directions 9.4 Conclusion References Part VI Last Thoughts 10 Concluding Remarks 10.1 Emerging Techniques 10.2 Future Work References