Design and Implementation of a Bandwidth Control Mechanism for Personal Wireless Routers

由於智慧型手機和平板的普遍,以及近年來各種行動連網裝置的出現,像是 穿戴式裝置及生理感測器等,透過這些裝置及相關硬體發展,能提供使用者更加 便利及多樣性的應用服務,但隨著個人擁有的裝置數量增加,每個行動裝置競爭 有限的網路資源,造成使用者操作不流暢或應用服務中斷,因此,如何有效控制 有限的頻寬資源,改善使用者體驗是未來使用情境重要的議題之一。 本論文中,我們討論以個人行動路由器(Personal wireless router)作為所有行動 裝置連網核心為架構,讓行動裝置透過短距無線通訊連結到路由器,再透過路由 器經行動網路上網,研究在個人行動路由器上實作頻寬控制機制,以改善服務品 質(QoS)相關的議題。我們先進行一連串實作的評估,從實作機制所使用的作業系 統和對於機制開發的有利元素,到頻寬控制演算法、利用封包探測(Packet probing) 的頻寬測量技術,並分析本論文最後使用的技術介紹及實測。 我們設計一個自動化的頻寬控制機制,並實作在 Android 智慧型裝置上。此機 制可分為三大部分:事件觸發器、頻寬測量,和頻寬控制。透過事件觸發器可以 偵測各行動裝置的使用狀況和相關資訊,並回傳給個人行動路由器,借此動態地 啟動、中斷或修改機制參數。在機制啟動下,利用 WBest 進行基於封包對探測之 頻寬測量,接著透過測量結果、即時狀態和服務品質資訊表,計算出適當的參數, 使用 Hierarchical Token Bucket 演算法實作頻寬控制。 本論文經由實際測量,利用同一個機制,在數個環境及操作情形下,並且與 不使用機制的原始設定做比較,由測量結果可得知,本論文提出的機制在描述的 情形下,皆能有效地改善使用者體驗及使應用較正常地運作,且不會造成裝置本 身過多的運算負擔。Due to the development of smartphones and tablets, the population of smart handheld grows up with a numerous speed in the passing years. Many other mobile Internet devices, such as wearable devices and sensors, have also appeared constantly. The number of mobile Internet devices per person is observed to follow an ascending trend, and various applications are under development. One of the current important issues is how to use the bandwidth resources more effectively to improve user experience when all devices compete for the scarce resource. In this thesis, we focus on the network architecture with a personal wireless router as the center of personal devices that offers Internet access over a wireless small area network. We aim to resolve related QoS problems by implementing bandwidth control mechanism on the personal wireless router. We made careful assessments and tests on key items of this architecture, including operating system on mobile devices, technologies of traffic control, bandwidth estimation and some measurement tools. The concept, methods and setup we choose to use in this thesis will be described, and testing results of every component are illustrated. We designed a mechanism called Dynamic Bandwidth Sharing Mechanism for Multiple Devices (MD-DBS) and implemented it on Android devices. MD-DBS consists of three major components: Event Triggers, Bandwidth Estimation and Traffic Control. The Event Triggers detect specific events of each mobile device and deliver necessary information to the router. The router can start or terminate the mechanism, and even reset the parameters dynamically, then it uses Bandwidth Estimation based on modified WBest to measure the available bandwidth. WBest was the chosen probing technology based on the packet pairs approach. The outputs of Bandwidth Estimation, instant information and QoS Tables were used for calculating proper parameters for controlling the bandwidth. Finally, we use the proposed Traffic Control mechanism based on the Hierarchical Token Bucket algorithm to effectively control the bandwidth. The proposed mechanism, MD-DBS, was evaluated under several usage scenarios and environments. We also compared the results with those under the default setting. By measurement results, we could observe that MD-DBS could improve user experiences and make applications work better. In addition, there was almost no significant computation overhead for devices using MD-DBS.口試委員會審定書 # 誌謝 i 摘要 iii ABSTRACT iv CONTENTS vi LIST OF FIGURES ix LIST OF TABLES xii Chapter 1 Introduction 1 1.1 Motivation 1 1.2 Problem Description and Research Goal 1 1.3 Related Work 2 1.4 Organization of the Thesis 4 Chapter 2 The Considered Network Architecture and Applications 5 2.1 Overview of Related Wireless Networks 5 2.1.1 Mobile Broadband Network 5 2.1.2 Small Area Networks 9 2.2 Network Architecture 11 2.3 Related Applications 13 2.3.1 Body Sensor Networks 13 2.3.2 Vehicle Networks 15 Chapter 3 Assessment of Alternatives and Implementation Environment 17 3.1 Operating System on Mobile Devices 17 3.1.1 Overview of Embedded Operating System 17 3.1.2 Android Architecture 18 3.1.3 Issue of Android Version 20 3.1.4 Useful Development Components 20 3.1.5 Command Line Interface in Android 22 3.2 Traffic Control 23 3.2.1 Overview of Linux Traffic Control 23 3.2.2 HTB Queueing Discipline 26 3.2.3 Setup of Traffic Control in Android 28 3.2.4 Testing of HTB 29 3.3 Bandwidth Estimation 32 3.3.1 Techniques of Bandwidth Estimation 32 3.3.2 The Wireless Bandwidth Estimation Tool - WBest 33 3.3.3 Setup of Bandwidth Measurement 35 3.3.4 Testing of Bandwidth Estimation 37 3.4 Measurement Tools and Methods 39 3.4.1 Network Metrics and Tools 39 3.4.2 Computation Overhead 42 Chapter 4 The Bandwidth Control Mechanism 44 4.1 Overview of MD-DBS Mechanism 44 4.2 Procedure of MD-DBS 46 4.3 The Procedure of Android Bandwidth Control Applications 50 Chapter 5 Measurement Results 53 5.1 Measurement Environment 53 5.2 Measurement Results 54 5.2.1 Bandwidth Control with Fluctuating Network Capacity 55 5.2.2 The Scenario of Changing Primary Devices 58 5.2.3 Ensuring QoS for a Specific Application 60 5.2.4 A VoIP Application over the Real Cellular Network 69 5.2.5 Comparison of CPU Utilization 72 Chapter 6 Conclusion and Future Work 74 6.1 Conclusions 74 6.2 Future Work 74 REFERENCE 7