隨著行動資料流量不斷增加，現今行動電信網路可能無法支撐未來龐大的網路流量。因此，我們審查邊緣計算，並研究ETSI基於此概念下，制定的行動邊緣計算(MEC) 規格，期望能解決未來行動電信網路面臨的困境。我們考慮影音串流的需求在未來也會大量增加，因此提出基於ETSI MEC的視訊串流服務，此服務包括用戶端的應用以及MEC端的應用。用戶端應用提供使用者搜尋與播放影片的功能，而MEC應用則提供用戶端應用快取管理和自適性影片畫質調整的服務。我們提出兩種機器學習的模型，分別用於快取管理來預測熱門影片，以及自適性影片畫質預測未來用戶端網路狀況。實驗結果顯示我們的快取置換機制的平均命中率，分別優於另外兩種方法15%和10%，而預測未來用戶端網路狀況的準確率能高達80%以上。

As the mobile data traffic keeps increasing, current mobile networks may become overloaded. Therefore, we survey the edge computing and study the specification of ETSI MEC. Considering that the requirement of video streaming is increasing, we propose an implementation for the ETSI MEC-based video streaming services, including the User App and the MEC App. The User App provides the function of searching and playing the videos for the end-users, while the MEC App provides the service of cache management and video quality adaptation for the User App. We propose two machine learning models for predicting popular videos and radio network condition which are used in cache management and video quality adaptation correspondingly. We compare our cache replacement strategy with the other two strategies and show an increase in the average hit ratio up to 15% and 10%, respectively. We also show that we can predict the future radio network condition with an accuracy of more than 80%.

Abstract i
Contents ii
List of Figures iii
List of Tables iv
1 Introduction 1
2 Overview of ETSI MEC 6
3 System Model 9
3.1 User App 10
3.1.1 User function 10
3.1.2 CQI reporting 10
3.2 MEC App 11
3.2.1 Cache management 11
3.2.2 Video quality adaptation 11
4 Implementation of User App and MEC App 12
4.1 User App 12
4.1.1 Search 13
4.1.2 Play 15
4.1.3 CQI reporting 15
4.2 MEC App 16
4.2.1 Cache management 16
4.2.2 Video quality adaptation 18
5 Prediction Model Based on Machine Learning Techniques 21
5.1 Popular videos prediction 21
5.2 CQI values prediction 22
6 Experiment Results 24
6.1 Popular videos prediction 24
6.2 CQI values prediction 27
7 Conclusion 29
Bibliography 30

[1] Cisco, “Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update, 2016-2021.” White paper, June 2017.
[2] Shi, Weisong, et al. “Edge computing: Vision and challenges.” IEEE Internet of Things Journal 3.5 (2016): 637-646.
[3] Patel, Milan, et al. “Mobile-edge computing introductory technical white paper.” White Paper, Mobile-edge Computing (MEC) industry initiative (2014).
[4] ETSI GS MEC 004: Mobile-Edge Computing (MEC); Service Scenarios.
[5] Abdelkrim, Emira Ben, et al. “A Hybrid Regression Model for Video Popularity-Based Cache Replacement in Content Delivery Networks.” IEEE Global Communications Conference (GLOBECOM), 2016.
[6] Liu, Yao, et al. “A comparative study of android and iOS for accessing internet streaming services.” International Conference on Passive and Active Network Measurement. Springer, Berlin, Heidelberg, 2013. J. Clerk Maxwell, A Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68-73.
[7] https://developers.google.com/youtube/v3/docs/. Retrieved Auguet, 2017.
[8] 3GPP TS136.213 version 13.0.0 Release 13. “LTE; Evolved Universal Terrestrial Radio Access (E-UTRA); Physical layer procedures.” (2016).
[9] McFarland, Patrick. http://adterrasperaspera.com/blog/2010/05/24/approximate-youtube-bitrates. Retrieved August, 2013.
[10] Zink, Michael, et al. “Watch global, cache local: YouTube network traffic at a campus network-measurements and implications.” Computer Science Department Faculty Publication Series (2008): 177.
[11] http://etd.lib.nctu.edu.tw/cgi-bin/gs32/hugsweb.cgi?o=dnthucdr&s=id=%22GH02103065524%22.&searchmode=basic. Retrieved Auguet, 2017.
[12] 3GPP TS136.101 version 14.3.0 Release 14. “User Equipment (UE) Radio Transmission and Reception.” (2017).
[13] Botta, Alessio, et al. “Internet Streaming and Network Neutrality: Comparing the Performance of Video Hosting Services.” ICISSP. 2016.