The 7th Niuyun Special Forum of the Cloud Conference was held in Hangzhou Cloud Town on September 20th, with the theme of “When Cloud meets AI”, centering on two key words of “Cloud” and “artificial intelligence”, many big names in the industry were invited to give wonderful speeches for everyone. Dai Wenjun, vice president of Qiuniuyun Technology, made a share entitled “How to use edge computing + edge storage to create a new generation of intelligent video cloud” at the meeting.
The following is a shorthand transcript of the speech. 
Good afternoon, everyone. Before we start, we should clarify that although the host and many friends say that edge computing is hot now, but in fact today, edge computing in the real sense of the ground is not too much, the application scenarios brought by the industrialization of edge computing have not started. The core of edge computing, IN my opinion, is the development of AI and big data.
So what are the areas where edge storage and edge computing can play a role? Today’s topic is Cloud meets AI. We serve a large number of customers from live broadcast, short video to online education. In the past year or so, a new big customer group has been added to our entire customer system, such as smart medical care and smart home. That’s a big change. 
In the industrial Internet, we need to be more “sophisticated” with video. Cloud meets AI in this space presents four new challenges. 
The first is more stable communication quality. For example, when we are watching a video, the Internet is not good, so we get stuck for a while. But we’re in surgery, and when it comes to forensics, that kind of thing can be a real problem. Therefore, this is the first challenge for video cloud in my opinion, and the most important challenge to be solved.
The second is lower latency. In the old days, when we were on TV or live, we could go live in seconds, with a delay of less than three seconds. In the new scenarios, such as new retail or wisdom park, there is a customer took one look after your shop goods, you need to quickly identify his related information, such as is our old customer, he is a what kind of purchase behavior and buying habits, this scenario delay more than four seconds, three seconds or maybe the customer left, We can’t make a connection. In our intelligent processing systems, lower latency will lead to new breakthroughs.
The third is greater demand for resources. In our life now, there are cameras everywhere, schools, parks, shopping malls and other scenes, it can be said that cameras are everywhere. But the storage of this data, it’s going to be a huge volume. At the same time, due to the relationship of log recording, its volume is much larger than the video, picture we imagined. Because the customer access log is accumulated over time, it requires greater resource requirements.
But I don’t think this part is the most important challenge, because there are technologies to overcome it and resources to solve it. In the last stage of the Internet, when we want to do picture applications, we will do a lot of picture processing related work, when we do video, we will do short video SDK and beauty, yellow and other functions. Whether it’s through SDK, APP, or even SAAS, address related intelligence needs. But who has the ability or the opportunity to solve the problems in smart medicine, smart home and smart parks? No one field can solve its real problems by copying. Until the deep-seated problems in this industry are solved, the development of the whole industry, including AI, will be delayed in the future. So I think the fourth challenge is deep industry convergence.
I mentioned resource issues, latency and communication issues, but at the end of the day, I think it’s network.
Let’s start with edge storage, and edge storage is a product that you rarely hear about. The industry calls it edge computing, and as long as the host has a hard drive, it counts as storage. But this situation only meets the needs of resources, not the needs of our technical services, which is the biggest problem.
So how do you solve a problem like that? In today’s cloud storage, if the nodes are all over the CDN nodes, can our protocol not change at all? This is what we pursue today. Therefore, the first point is to use edge storage to solve the big data storage, only in this way, no matter we carry out the analysis of big data or visual intelligent AI, we have a foundation.
With the storage problem solved, does it require everyone to do basic computing skills? We believe that not only do we need to move our customers’ applications to the edge, but we also need to help them with the updating of their applications and the deployment of those applications. It also needs to scale properly.
So when computing and storage are tied together, problems like cameras or health care or campuses can be solved much better. But we don’t just want to put storage power or computing power on the edge, we want to put the whole system on the edge. Give our customers apis, and they just tell us where to go, how much scheduling power we need, and how much storage power we need.
On the other hand, I think P2P is still an important direction for future development. How to connect point to point more quickly, how to better optimize the uplink network, are not the focus of our entire industry today. However, IN my opinion, to further meet the development of AI and big data, upstream optimization should be the focus of the technology. Of course, people will say that 5G has come, big data transmission is not a problem. But when does it land? Are 4G operators so determined to push out the infrastructure to get into 5G? I think there’s a process. It’s not going to be that fast, even with 5G, I think the upload volume will be bigger and our technical requirements will be stronger. 
Our entire edge storage and edge computing is called “stars”. And the reason for that is because we think that all the computational power is distributed and connected. This connection is not done by the customer or himself, but by treating it as a big galaxy, allocating and distributing all resources.
The video is a look at some of the changes we’ve seen in the industry and some of the challenges that customer change has brought. The core point of smart cloud is not distribution, but our production side. Side, edge, cloud computing resources cloud platform, we will now use some edge storage, storage with three copies or two copy it down, the ability to calculate the flow section, the concentration of video, video structured, put all these application ability in the edge, so that we can adjust our API, we can have the ability of video processing.
Let’s talk about edge storage and edge computing using two case scenarios.
First of all, this is our client’s previous architecture, and the upgraded business architecture is hosted by edge storage. When we moved the storage to the edge, we were able to meet the client’s huge volume requirements. Because for storage, camera, uplink is free, but each resource is very limited in this, if not with another piece to balance the cost, the cost is much higher than the storage cost. We put storage on the edge, we can meet the next 10 years, 20 years is no problem, now don’t worry about cloud service providers bandwidth capacity.
Further, for IP Camera devices, our edge storage capacity and streaming media capability can fully support 7×24 hours of upload. So the overall cost of the camera, and the cost of running it, would be a huge drop. Finally become edge storage to undertake SD card content and NVR storage function. The launch of our edge storage is an important channel for the camera to move towards cloud storage and cash in the future.
Let’s talk about another case. This is a client that we are close to delivering. He is an intelligent studio. 
Intelligent studio based on AI is a complex system, how to manage its basic IT resources? What about basic storage? What about basic calculations? Because the studio is all over the country, there are dozens of stores in a city, there are hundreds, thousands of stores nationwide. Is it possible to only care about his application upgrade, do not care about how the machine system is maintained, how to schedule behind. We take the container, and we move the container to the edge, and the center delivers the edge container to update its business structure. It used to take five days to launch a studio project, but now it takes one to two hours, which is a huge performance improvement. The products we serve for studios can also serve other industries, such as distributed business of real estate companies and chain stores.
We’ve solved the storage problem, we’ve solved the computing problem, but what is it that we haven’t completely solved at the core? Application landing issues. How can we support smart medicine, smart car scenarios? In the larger ecosystem, any family can only make a fan side.
Edge storage and edge computing can be implemented in our public and private clouds. But what is its premise? Everyone follows certain standards. The standard is currently less restrictive for applications, namely containers. I think this is a problem that needs to be solved in the future. This does not prevent us from taking the first step to go out today. Let’s go to the big circle to share. I also hope that more people will walk on our basic video architecture in the future, so that the whole Cloud will meet the AI of society. 
