基於邊緣運算架構的直播服務與即時動態轉碼決策

Abstract

影片串流與直播服務已經成為網際網路上的熱門應用。現在主流的直播服務為使用自適性影音串流的技術以避免因網路波動造成使用者品質下降另一方面，隨著5G 網路的發展，由ETSI 所主導的邊緣運算正大幅改變整個網路架構。各種應用程式如AR，VR，及影片、實況串流對於延遲以及網路頻寬的要求越來越高，這些結合邊緣運算的平台都可以大幅得到改進。 本論文提出了一個基於邊緣運算架構的直播服務系統。基於直播服務最大的瓶頸在於穩定的上傳品質，利用邊緣運算的架構可以大幅提升上傳、下載的網路品質，並且將利用將轉碼放在邊緣運算可以減少後端網路的流量。同時為了避免過多的負擔造成整體使用者體驗下降，我們提出了即時動態轉碼決策的方式在較高負擔時同時維持一定的服務品質。 實驗平台上的評估結果顯示，使用我們的機制可以有效的提升資源利用效率，並提供較好的網路品質同時提升整體的使用者體驗。總體而言，我們提出的架構以及管理機制揭示了基於邊緣運算優化直播服務的可能性。

Video streaming and live streaming has become popular applications running over the Internet. Adaptive video streaming is now the trend when conducting live streaming service, and DASH and HLS are the mainstreaming in this field. On the other hand, with the development of 5G network, Multiaccess Edge Computing (MEC) led by ETSI is greatly changing the network architecture. Applications such as AR, VR, video streaming and live streaming which demands lower latency and higher network bandwidth may be improved by deployed on MEC platform. This thesis proposed a live streaming system based on MEC architecture and platform. Because the bottleneck of a live streaming service is a stable and large uplink bandwidth, use the MEC architecture can dramatically enhance the uplink and downlink channel quality, and transcoding on the edge can lower the backhaul traffic. To avoid too much computing load causing degradation of QoE, we proposed a realtime dynamic transcoding decision mechanism to provide an acceptable service quality under the heavy load situation. Evaluate result on our testbed shows that our mechanism can enhance resource utilization and provide a better channel quality to enhance overall QoE. All in all, our proposed architecture and management mechanism reveals the possibilities to optimize the live streaming service.