LTE 網路是 4G 通訊的主流技術，在未來的 5G 架構裡，它仍將扮演重要角色以支援更多的裝置應用。在未來的全球行動裝置流量中，影音串流類型資料仍會占大宗。要如何在現有技術架構中，更有效率的使用 LTE 頻譜資源是很重要的課題。隨著影片畫質越來越高， video server 採用自適性串流技術行之多年。目前的主流技術為基於 HTTP 、 TCP 的 DASH (Dynamic Adaptive Streaming over HTTP) 技術。若單純以 TCP 控制網路流量並無法最大化無線頻譜資源的使用。本研究以影音串流為應用，結合 ETSI 所提出之 MEC 架構，希望最大化資源使用效率。我們透過 EURECOM 所提出之 LTE eNB 平台 – OpenAirInterface (OAI) ，於其上完成 MEC 架構的自適性串流實驗。在第一部份裡，我們首先談到「Filter」模組，我們藉由更動 eNB 位於 MAC 層的程式碼，我們成功的模擬出 MEC 架構下， MEC server 與 video server 間參考了 CQI 的互動情形。在第二部分裡，我們談 OAI 本身的 resource block (RB) 排程演算法，我們改善了 OAI 多 UE 排程 (multiuser scheduling) 的資源分配問題，這會跟我們實驗的throughput有很大的關係。 關於實驗結果，在第一部份裡，我們會以固定資料速率作為對照組，以檢視我們自適性串流的實驗效果；在第二部分裡，我們會以未更動 RB 排程演算法的程式碼作為對照組，以檢視效能。藉由這兩個部分的程式碼更動，我們也完成了MEC 架構的驗證。

LTE is the leading technology for 4G mobile communications, and is expected to support an even wider range of devices in the 5G era. The video traffic will become majority of global mobile data traffic in the future. It’s a critical issue to increase the spectral efficiency under the existing framework. As high-definition video be more prevalent, video servers have been actively using adaptive bitrate streaming technology for many years. DASH (Dynamic Adaptive Streaming over HTTP) is the standard for adaptive bitrate streaming technology, which uses TCP as the transport protocol. As the limit of TCP protocol, it’s hard to best utilize the bandwidth usage within a wireless environment. By using video streaming as our application, we studied on the MEC (Multi-access Edge Computing) framework proposed by ETSI to make better use of resource. We adopted EURECOM OpenAirInterface (OAI), an LTE eNB software as our testbed to implement our MEC-based adaptive video transmission study. By modifying the code in the MAC layer of eNB, we simulated the interaction between MEC server and video server under MEC framework. We will first talk about “the Filter” in Section V’s subsection A. We design and implement a “filter” module in the Medium Access Control (MAC) layer over the OAI platform, which can reference Channel Quality Indicator (CQI) as a parameter to decide which bitrate type of data should be passed to the MAC Packet Scheduler for real resource allocation. We than talk about the resource block (RB) scheduling algorithm of OAI, which plays an important role in the throughput performance during our experiment in Section V’s subsection B. On the result of our implementations, in part I, we used constant-bitrate as our control groups, the testbed indicates that, consistent with our expectations, our mechanism can significantly benefit from lower latency with a reasonable throughput; in part II, by comparing with the code before our RB scheduling algorithm modification, we also showed that the latency and throughput have been greatly improved. Based on these two-part implementations, we validated the MEC architecture.

Contents
摘要 I
Abstract II
Contents IV
List of Figures V
List of Tables VI
I. Introduction 1
II. MEC-based Adaptive Video Transmission 3
III. LTE Scheduling 5
A. The Radio Channels 5
B. CQI MAC Scheduling 6
C. HARQ 6
D. Video Quality and Synchronization 7
IV. The MEC Scenario 9
V. Realization of an LTE Testbed for MEC-Based Video Streaming Using EURECOM OpenAirInterface 11
A. The Filter Module 11
B. OAI Scheduling 13
VI. Implementation 15
A. Pseudo Code for the Filter Module 15
1) The Filter Operations 15
2) The Filer Pseudocodes 16
B. Code Modification for OAI RB Scheduling 18
VII. Experimental Results 28
A. Experimental Environment 28
a) SNR Models 28
b) System Parameters 29
B. Performance for the Filter Module 29
a) Latency vs. Fading Profile Model 30
b) Throughput vs Fading Profile Model 30
C. Performance for OAI RB Scheduling Modification 31
a) Latency vs. Fading Profile Model 31
b) Throughput vs Fading Profile Model 31
VIII. Conclusion 32
References 33
 
List of Figures
Fig. 1. Intelligent Video Acceleration 3
Fig. 2. MAC multiplexing. 6
Fig. 3. HARQ 7
Fig. 4. Video qualities. 7
Fig. 5. Concept of Decision Time Interval (DTI). 8
Fig. 6. MEC-enabled LTE channel-aware video streaming. 10
Fig. 7. The Filter in the testbed. 12
Fig. 8. The filter interface. 12
Fig. 9. The synchronization concept. 13
Fig. 10. The filter operations. 15
Fig. 11. Pseudocode for choosing a suitable DTCH based on CQI. 17
Fig. 12. Pseudocode for adjusting the transport block size. 17
Fig. 13. Pseudocode for the filter concept. 18
Fig. 14. OAI RB scheduling algorithm. 18
Fig. 15. Store Downlink Share Channel Buffer 19
Fig. 16. Assign Required RBs 20
Fig. 17. Sort UEs 21
Fig. 18. Calculate Average BW RB 22
Fig. 19. Hypothetical Assignment 23
Fig. 20. RB Pass Mechanism 25
Fig. 21. Calculate Minimum RB (Original) 26
Fig. 22. Calculate Minimum RB (Modified) 26
Fig. 23. RB Allocation in 2 Rounds 27
Fig. 24. RB Allocation in 2 Rounds (for the control group) 27
Fig. 25. Latency vs. fading profile model. 30
Fig. 26. Throughput vs. fading profile model. 30
Fig. 27. Latency vs. fading profile model. 31
Fig. 28. Throughput vs. fading profile model. 31

 
List of Tables
Table 1. Notations 16
Table 2. Fading Profiles 28
Table 3. System Parameters 29

[1] Cisco VNI Mobile Forecast (2015 – 2020) http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/mobile-white-paper-c11-520862.html
[2] MPEG. "MPEG ratifies its draft standard for DASH". 2011-12-02. Retrieved 2012-08-26.fig
[3] Mobile-Edge Computing (MEC); Service Scenarios. ETSI GS MEC-IEG 004 V1.1.1, November 2015.
[4] OpenAirInterface website, url: http://openairinterface.eurecom.fr/
[5] 3GPP TS 36.101 Annex B.
[6] Østerbø, Olav. "Scheduling and capacity estimation in LTE." Proceedings of the 23rd International Teletraffic Congress. International Teletraffic Congress, 2011.