This article is an abstract of a paper originally published in “Broadcast and Television Technology” Issue 5, 2022.
With the development of 5G networks and the rise of live broadcasting services, edge computing will play an important role in live broadcast scenarios. This article starts with the concept of 5G edge computing and studies its application in low-latency live broadcasts, focusing on the construction of application platforms and experimental results, which confirm that 5G edge computing can reduce latency in live broadcast services.
Authors: Zhou Xiang, Xiao Jingtong, Li Xunchun
National Radio and Television Administration, Institute of Broadcasting Science
First Author Introduction:
Zhou Xiang (1993—), male, assistant engineer at the Institute of Broadcasting Science, National Radio and Television Administration, master’s degree. Mainly engaged in research on 5G, audio and video transmission, etc.
1 Overview of 5G Edge Computing
Support for edge computing is a key technology in 5G systems, considered from the outset of 5G system design to provide efficient and flexible support for edge computing. 5G edge computing meets low-latency transmission through locally available computing resources while saving backbone bandwidth resources.
2 Live Broadcast Architecture Based on 5G Edge Computing
2.1 Platform Introduction
This article builds a 5G edge computing live broadcast system based on the laboratory’s 5G platform, as shown in Figure 1.
On the terminal side, the experimental network uses commercial terminals as access devices for testing, such as 5G-enabled smartphones or 5G-enabled CPE devices. During live broadcast testing, streaming can be done by installing an app on the phone or by connecting the CPE (Customer Premise Equipment) to a computer and using proprietary streaming software. Playback can also be done on either the mobile side or the computer side, using the WebRTC protocol for direct playback in the browser, which will be detailed below. This experiment uses Huawei’s CPE PRO2 for testing, as this device supports the 700M frequency band for broadcasting.
On the base station side, a 5G base station based on Software-Defined Radio (SDR) is used. This base station is based on an x86 system and uses specialized physical boards for radio frequency output. The base station supports carrier aggregation and MIMO, allowing multiple terminal access, and this experiment uses the 700M frequency band for testing.
On the core network side, considering the need for distributed cloud deployment of the UPF network element, an open-source core network approach is used for cloud deployment, separating the user control plane and data plane of the core network. The control plane and UPF1 of the core network are deployed on a remote public cloud platform, while a remote cloud live broadcast system is deployed in the same data center as UPF1. The local UPF2 is deployed on a local server directly connected to the base station and connected to the local MEC service to achieve low latency edge computing.
On the bearer side, considering the actual situation in the laboratory, an operator’s dedicated line in the laboratory is used as the public network between the remote cloud and the remote core network and UPF. The base station is directly connected to the local UPF through a switch.
At the service layer, the live broadcast streaming uses the Real-Time Messaging Protocol (RTMP), a proprietary application layer protocol designed by Adobe for audio and video data transmission services between Flash players and servers. The RTMP protocol is the basic live streaming protocol used by major cloud vendors, characterized by stability, strong compatibility, and high penetration. With the development of the domestic live broadcast industry and the arrival of the 5G era, it is commonly used in scenarios such as streaming media live broadcasts and is one of the mainstream streaming methods today.
For pulling stream technology, the WebRTC protocol, which currently supports low latency well, is used. WebRTC enables web-based video live broadcasting, aiming to provide real-time communication capabilities through simple scripting languages in browsers. This protocol can achieve real-time delays of about 500ms, with actual measured delays around 700ms, depending on network conditions. In addition to low latency, the WebRTC protocol can natively support browser playback without the need to install any plugins. It is widely used in scenarios such as video conferencing and live broadcasting.
For edge diversion technology, this experiment uses the DNN scheme. As mentioned above, the SMF is responsible for UPF selection. The terminal configures a dedicated DNN and signs a dedicated DNN on the Unified Data Management (UDM) of the core network. Users initiate session establishment requests through the dedicated DNN, and when the SMF selects the UPF, it chooses the destination edge UPF based on the dedicated DNN provided by the 5G terminal, completing the establishment of the edge PDU session to connect to the MEC platform associated with the edge UPF.
Thus, in this experiment, by configuring different DNN information for UPF1 and UPF2 in the SMF, UPF selection is conducted. Specifically, by configuring different DNNs on the terminal, the SMF can select the appropriate UPF for the terminal based on the DNN information, achieving the switch from remote UPF to edge UPF, thereby enabling the terminal to switch between remote cloud and edge cloud. The DNN scheme has low requirements for terminals and networks, making it possible to quickly launch MEC services during the early stage of 5G commercial use. However, as 5G services develop, allocating independent DNNs for each MEC customer will pose a significant challenge to core network devices, especially in terms of the number of UPFs supporting DNNs.
In this experiment, UPF1 and UPF2 are distributed at different geographical locations. To more realistically simulate the latency situation of the remote cloud, this article selects a remote public cloud located in Suzhou as the location for UPF1, with an average two-way latency of 73ms from the terminal to the remote cloud. UPF2 is deployed in the laboratory and directly connected to the base station, with an average two-way latency of 32ms from the terminal to the edge MEC. Due to the current R15 standard of the laboratory’s 5G base station and the performance limitations of SDR, the one-way latency on the air interface is about 15ms. Thus, the latency from the terminal to the edge MEC is primarily caused by the wireless side’s air interface latency.
2.2 Experimental Process and Results
Before conducting the live broadcast experiment, first connect the terminal to the 5G network, as shown in Figure 2. It can be seen that Huawei’s 5G CPE device has successfully connected to the 5G experimental network and displays the China Broadcasting logo. Next, one terminal is used for streaming, while another terminal is used for pulling the stream, calculating the live broadcast latency by comparing the time difference between the streaming side and the pulling side.
Below is an introduction to the specific method for calculating latency. First, prepare a millisecond-level clock display, capture it with the streaming terminal, and conduct pull-stream playback with the pulling terminal. Then, take pictures of the clock displayed on the pulling terminal and the captured clock simultaneously, ensuring that the time scales are in the same frame. After capturing the photos, extract the time from both, and subtract the timestamps in the image to obtain the end-to-end latency value for the live broadcast. For convenience, this experiment uses screenshots instead of photos. The experimental results are shown in Figure 3, where the left clock is a continuously changing clock page, and the right side is the clock page displayed on the pulling terminal. By subtracting the clock scale value on the right from that on the left, the latency for this live broadcast can be calculated, as shown in Figure 3, where the latency is 591–50=541ms.
This experiment calculates the average value through multiple calculations for the statistical results of latency in remote cloud and edge cloud live broadcasts. When using the remote cloud for live broadcast services, the average latency is 704ms. When using the edge cloud for live broadcast services, the average latency is 654ms.
The purpose of this experiment is to verify whether the low latency brought by the edge cloud can effectively reduce the latency of live broadcast services. The above experimental results show that when the latency difference between the terminal to the edge cloud and remote cloud is 40ms, the resulting live broadcast latency is about 50ms. When using the edge cloud for live broadcast services, it can reduce the latency by 50ms compared to using the remote public cloud.
Additionally, with the commercial use of 5G air interface devices supporting R16 and R17, the wireless side latency will further decrease to a few milliseconds. Moreover, as live broadcast protocols continue to evolve, some live broadcast cloud service providers adopt self-developed proprietary protocols, which can significantly reduce live broadcast service latency, potentially achieving end-to-end latencies of several tens of milliseconds. Thus, the latency reduction brought by 5G edge computing to the live broadcast industry will become even more meaningful.
3 Conclusion
This article first provides an overview of 5G edge computing technology and, based on this, introduces the research and application of 5G edge computing in low-latency live broadcasts, with detailed descriptions of network architecture, live broadcast technology, and more. By comparing live broadcasts from 5G edge clouds and remote clouds, it demonstrates that edge computing can reduce latency in live broadcast services, showcasing the role of 5G edge computing in live broadcast services. In the next step, with the construction of the broadcasting 5G network and the gradual deployment of broadcasting 5G edge clouds, network audiovisual services will get closer to users, undoubtedly further promoting the landing and development of 5G high-definition video services, bringing users a higher, newer, and better audiovisual experience.
This article is funded by the Basic Research Business Fund Project of the Institute of Broadcasting Science, National Radio and Television Administration, 2021, titled “Research on Edge Cloud Technology for 5G High-definition Video” (Project Number: JBKY20210110).
END
References
[1] Song Xiaoshi, Yan Yan. Key Technologies of MEC Systems for 5G [J]. ZTE TECHNOLOGY JOURNAL, 2018.
[2] 3GPP. System Architecture for the 5G System [EB/OL]. [2019-12-22].
https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3144.
[3] Zhou Xiang. Research on Adaptive Transmission Mechanism of Mobile Edge Computing Based on Wireless Network Information [D]. Beijing University of Posts and Telecommunications, 2019.
[4] Chang Zheng. Research on Resource Optimization Based on Multi-access Mobile Edge Intelligent Computing [D]. Beijing University of Posts and Telecommunications, 2020.
[5] Tan Yiming. Research on User Business Optimization Based on Mobile Edge Computing [D]. Beijing University of Posts and Telecommunications, 2019.
[6] Zhou X, Chang Z, Sun C, et al. Demo abstract: Context-aware video streaming with q-learning for MEC-enabled cellular networks [C]//IEEE INFOCOM 2018-IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). IEEE, 2018.
[7] Chang Z, Zhou X, Wang Z, et al. Edge-assisted Adaptive Video Streaming with Deep Learning in Mobile Edge Networks [C]//2019 IEEE Wireless Communications and Networking Conference (WCNC). IEEE, 2019.
[8] Wang Lei, Zhang Zhichao. Shunting Technology and Deployment Methods of 5G UPF [EB/OL]. [2021-03-02].
https://zhuanlan.51cto.com/art/202103/647888.htm
[9] ETSI. MEC in 5G networks [EB/OL]. [2018-6-1]. https://www.etsi.org/images/files/ETSIWhitePapers/etsi_wp28_mec_in_5G_FINAL.pdf
“Broadcast and Television Technology”, “Guangdian Lieku” advertising operation and business cooperation agency:
Beijing Zhongguangxintong Cultural Media Co., Ltd.
Contact: Li Cong
Contact number: 18518221868
Please share this good article. If you have something to say, please leave a message.
