Abstract: Huawei cloud TechWave cloud native media services on the special day, huawei cloud cloud video media services products division director Lu Zhenyu speech “cloud native media network, upgrade the traditional, fu to the future”, why share the refactoring media network, has built based on huawei cloud cloud native technology media network can solve the current problem, And why cloud native is a must for future media networks.
This article is shared from huawei cloud community “Huawei cloud based on cloud native media network, and a new blockbuster”, the author of the original: audio and video butler.
On June 25, Huawei Cloud TechWave cloud native media Service Special Day was successfully held online. Huawei cloud audio and video services have been upgraded to media services. The upgraded media service supports media production, distribution, and application. On the special day, huawei cloud cloud video media services products division director Lu Zhenyu speech “cloud native media network, upgrade the traditional, fu to the future”, why share the refactoring media network, has built based on huawei cloud cloud native technology media network can solve the current problem, and why is cloud native media network will option in the future. The following contents are arranged and deleted according to the content of the speech.
Challenges and opportunities coexist: traditional networks have been unable to adapt to the surge of traffic and new business forms
In the last few years, we’ve seen global traffic, especially video traffic, increase 12-fold in five years with the advance of mobile Internet and 5G. We’re seeing a lot of 1080p, 4K, 8K, VR video, and the network is bursting with traffic and bandwidth.
Traditional CDN and live broadcast can well solve the distribution of on-demand video and traditional live broadcast video, but the application scenario is limited because of the delay of 3-5 seconds, which cannot support the two-way interaction and real-time interactive video distribution scenario. For example, ultra-low latency live broadcast, real-time audio and video, and some video surveillance scenarios, these services have higher requirements for media distribution capabilities.
In the scene of live broadcasting with goods, the host often starts to introduce the product, and the user side shows that the link is on. However, 5 seconds later, the user hears the host say that the countdown starts, and the link is on 5, 4, 3, 2 and 1. If the user clicks the link again at this time, he will find that the product may have been robbed. That’s because there’s a five-second delay between the broadcast and the text. In such a scene, it is necessary to broadcast with ultra-low delay within 1 second, which is not available in the current media distribution capability, and we call it “too late”.
Another scenario is “can’t afford it.” Everyone wants to get the content from video cameras to the cloud cheaply and quickly, because there’s a lot of AI capability and a lot of business innovation in the cloud that can make video more valuable. With the current model calculation: take an ordinary small park as an example, 100 cameras each bandwidth of 1 MB, is 100 MB, such a cloud broadband access cost will be more than 120,000 yuan a year, which is far from affordable. Due to the unaffordable cost, many parks can only store these video content offline. The problem is that these video content assets cannot be well realized and create more value.
Bid farewell to traditional networks: Cloud native media networks with converged media nodes, layered design and AI scheduling features are born to meet the needs
In order to solve such problems as “too late” and “unaffordable”, Huawei Cloud proposes the cloud-native media network. We believe that the cloud-native media network should have three characteristics: integrated media nodes; 3, 4, 7 layer hierarchical network transmission optimization; Intelligent scheduling of AI.
Converged media nodes In the past, when we had CDN services, there were CDN nodes, and when there were live broadcast services, there were live broadcast nodes. Each RTC and video surveillance service has its own node. These nodes cannot be shared or elastic, limiting service deployment.
The first thing Huawei cloud did was to transform 2500 nodes around the world in a unified way of cloud native and upgrade them into converged media edge nodes. Its characteristics are unified computing and storage resources at the bottom, and unified transformation and management are carried out on the top with the original IEF. In this way, CDN, live streaming, RTC, XR, transcoding and other services are dynamically deployed on the basis of cloud native capabilities.
Each service can be scheduled during the day and at night based on its own requirements to use uplink and downlink bandwidths and computing and storage resources. In this way, edge assets are activated, which greatly reduces service transmission latency and more importantly, service costs.
Layered design We use layered ideas for network transmission optimization: similar to TCP/IP to consider audio and video content, transmission on the network, it encountered packet loss, delay and different network access conditions under different parameters of the problem, we use layered ideas to solve, divided into 3 layers, 4 layers and 7 layers optimization. At layer 3, that is, IP layer, we put forward the sky route, which can sense the network in real time and optimize the forwarding and routing, so as to improve the delay of packet forwarding and improve the arrival rate of packets. Through actual measurement, we can reduce the forwarding delay of IP packets on IP network by 30% and increase the arrival rate of packets by 0.5 percentage points.
Also in the fourth layer, namely the transmission layer, the hQUIC protocol developed by Huawei is proposed, so that the upper application development does not have to perceive what kind of network media the lower layer is running on, whether it is WiFi, 5G or other kinds of network. It can accelerate the content of audio and video service, real-time message transmission service and future XR cloud game service respectively under different network conditions, and improve the efficiency of network transmission.
In layer 7, the problem of audio and video service transmission is solved in various business scenarios. For example, is it necessary to carry out some codec first and then transmit to improve efficiency? Whether to carry out some adaptive codec parameters, whether there is a better experience in the middle of transmission…… This is all done on the seventh floor. We optimize layers 3, 4 and 7 so that the media network can provide a very consistent experience for different audio and video content on different networks.
AI scheduling we put forward a Mesh scheduling engine, which is the result of our cooperation with the Top1 schools in China. It has two key features. The first feature is the unified scheduling of multiple services, which includes vod, live broadcast, RTC, monitoring, etc. We can complement and tune the experience and cost to realize the end-to-end unified scheduling capability of multiple services. Different services are scheduled uniformly by this scheduling engine.
The second feature is that different service SLAs adopt different scheduling strategies to support different customers’ business strategies. For example, for live streaming, we pay more attention to the back source rate, right? We take these indicators as an input to the scheduling engine. Among 2500 nodes, we always choose the optimal node for dynamic deployment, elastic scaling and optimal service.
For RTC such higher real-time, interactive stronger, experience index more demanding scenarios, we will put the user’s first frame length, number of caton, into the room with a success rate, end-to-end delay, to set of input parameters, such as scheduling engine, the RTC always provide a better user experience, lower time delay, power the customer’s business. In order to realize the scheduling capability, we implement Channel level scheduling in the scheduling system.
We conduct overall scheduling and optimization for video access stage, back source stage and the whole network according to a Mesh network, and introduce artificial intelligence to conduct self-learning and self-training for scheduling algorithms, and constantly confront each other to generate better scheduling strategies and algorithms. Through the unified upgrade of media edge network, our AI scheduling, and the optimization of network transmission layer 3, 4 and 7, our cloud native media network can solve the problem of “too late” and “unaffordable”.
Give you two examples, the first example is above such media network, we can live broadcast service upgraded to ultra low delay, now live with 3 ~ 5 seconds delay, we’ll live with RTC live upgraded to ultra low delay, delay is reduced to less than 800 milliseconds, but other indicators business experience, The first-screen lag is also on par with traditional livestreaming, and such services will help livestreaming, including e-commerce and education industries, make a big experience upgrade.
At the same time, if the client is willing to accept the transfer of its architecture station from live broadcasting to RTC, we can also directly provide RTC services on such a cloud native media network, so as to change the client’s transmission from TCP to UDP, bringing further experience upgrading and delay reduction.
The second example is the cloud native video surveillance service (VIS service), which is also based on the cloud native media network. It fully reuses the uplink traffic in the network and makes the access price of the camera more affordable. From me just now, for example, from 300 yuan a month, a month two digits, can make each camera operations cost greatly reduced, and prompt the user to camera generated content on the cloud, enjoy a lower cost of storage, and the cloud more innovation and the ability of AI, we can do that in the past users monthly cost of less than 10%.
New product: open, sharing, efficient and secure video access service, based on edge collaboration architecture, supporting multiple business scenarios
Today I am honored to launch Huawei’s VIS Service product, Video Ingestion Service. The video access service is an open, sharing, efficient, and secure video access service based on the edge collaboration architecture of huawei cloud. It supports multiple business scenarios and supports video cameras, similar to the access of national standard 28181 video protocol, RTMP video and FLV video. We will continue to add support for SRT and more intelligent cameras and video access protocols in the future.
At the same time, we will continue to expand the boundaries of service, upgrade the value of service. The content of video is back on the cloud. Only when it colliders with more industry ecology can it generate more innovation and bring more value.
Huawei cloud not only has self-developed algorithms, there are open algorithm mall, there are a lot of third-party algorithms, these algorithms are continuous evolution, infinite overlay. We believe that more value will be created by bringing more video to our cloud at a lower cost. Then we can watch a short video to see how Huawei cloud video access service can easily make video camera content on the cloud.
I am very glad to share with you the native media network of Huawei Yunyun. We introduced the challenges we met and how Huawei solved these problems by integrating media nodes, layer 3, layer 4 and layer 7 hierarchical optimization, and AI scheduling. We launched for ultra low delay live broadcast, VIS video access such products. Thank you.
Click to follow, the first time to learn about Huawei cloud fresh technology ~