🎂 Dacheng Tao & Jonathon Dixon Fireside Chat
2021 marks Cloudflare’s 11th birthday, and each day this week we will announce new products and host fascinating discussions with guests including product experts, customers, and industry peers.
During this session, Jonathon Dixon (VP & GM of Cloudflare APJC) will host our guest Dacheng Tao (SVP of JD.com) to discuss the evolution of JD Cloud and the techniques in the future.
Find all of our Birthday Week announcements and CFTV segments at the Birthday Week hub
So Dacheng, why don't you tell us a little bit around the evolution of JDCloud? Where the company spawned from, some of the aspirations of what you're doing today, and what you hope to achieve in the short to medium to long term across this market?
Okay, there's so many questions. Jonathon, thank you for your questions.
So I mean, first, explain so the inception of JDCloud or why why JD need its own cloud?
You know, that's a so the enormous demands from APP users inspired JD engineers and researchers to develop its own cloud.
We built a cloud platform that could support JD's business development, support trillion Chinese Yuan e-commerce transactions, and robustly handle the peak of Internet traffic from JD's annual sales events like 618 and W11.
The robustness of our cloud solution has enabled 100% of orders during both events to be fulfilled online through our own cloud.
Following the cloud evolution, services from both JD Logistics, JD Health have been deployed to the JD Cloud.
These vertical services have progressively built JD's industry know-how and developed JD's strength for enterprise cloud.
I think maybe regarding the initiative of JD's cloud, I think maybe I can commence by introducing the 618 event earlier this year.
That's a very typical event in JD. During this recent event, the net order value amounted to a record setting nearly 350 billion Chinese Yuan.
We have to understand that this most customer experience during the event shopping festival was fully supported by our powerful cloud computing platform.
So the JD Cloud was born of necessity in dealing with the high demands of JD's larger scale app usage and was the first used to satisfy JD's internal app requirements.
We created a cloud platform capable of supporting JD's business development, supporting enormous e -commerce transactions, and robustly handle the peak of Internet traffic from JD's sales events, ultimately enabling the fulfillment of all orders during the event to be completed on the cloud.
The modern JD enterprise cloud supports the world's largest network with the ability to manage over 5 million SKUs for commodities.
At the heart of this network is a massive warehousing system distribution center and an end-user service framework.
The capacity afforded by this network has enabled JD to service 5 million customers, achieve one-day deliveries, and national service coverage.
Empowering all this is a process in power and advanced system intelligence of incomprehensible magnitude.
The robust handling of the annual sales events is evidence of continued evolutionary improvements of JD enterprise cloud capabilities, cementing its position as a technological cornerstone of JD.
JD and JD Cloud's realization has proven that cloud computing is becoming a key driver in enterprise digital transformation.
With the growth influence of cloud-based services, many new companies are joining the cloud computing sector every year.
According to Gartner's 2021 report, JD Cloud ranks fifth as a domestic ISA provider, becoming a top-tier provider in China.
According to our insights within the industry, no cloud service provider could completely satisfy all the demands from various enterprise applications.
Therefore, the digital transformation within the industry has led to the rising prominence of hybrid cloud solutions.
The closed-basic cloud model in use cannot satisfy requirements of high-level enterprise users.
It is becoming increasingly difficult to manage enterprise applications requiring inter -cloud service migration.
Based on our own insights to the industry, JD Cloud was conceived to serve the needs of the industry.
Driving industrial transformation that is specifically tailored to the requirements of enterprises and promoting business growth through digital transformation are the main objectives of enterprise cloud services.
Providing tailored services will be the main application of cloud computing and the development direction of the JD Enterprise Cloud.
So JD Cloud is also an industrial cloud.
JD is deeply engaged in the industrial Internet. We have accumulated first-hand experience in many fields, such as retail, logistics, and supply chain finance.
And we share our experience with the industry through our JD Cloud.
For example, the JD Cloud helped a co-group of companies to build an intelligent digital transportation and marketing platform.
Managing a digital transportation and marketing platform is a core skill employed by JD in its daily operations.
The ability to move in the same direction helps customers smoothly access the upstream and downstream supply chains, comprehensively improve the current work process over the origin.
This has led to the reduction of process time by 40% and increase the co-transportation efficiency by 30%.
JD Enterprise Cloud also offers larger scale coverage to customers across different industries.
For example, the AI-based human-computer interaction experience accumulated in the field of retail customer service is being applied to various urban service scenarios and has achieved good results that satisfy urban residents.
To date, JD Cloud has served customers in nearly 60 cities, including thousands of larger enterprises, more than 700 financial institutions, and millions of small and medium businesses.
The accumulated customer service experience offers us an advantage and enables us to continue delivering a high -standard service to our customers.
Right, thank you.
Wow, that's a really great summary. Obviously, a lot to digest there. You know, it's remiss of me, Dacheng, not to sort of get you to introduce yourself as well, because you only just started with JD.com in March, and you've got a rich sort of background, which probably paints a picture to a lot of the things you just said there.
So before we go to the next question, you just talk a little bit about your background prior to joining JD.com.
All right. Okay, so thank you, Jonathan.
Yes, I commenced my position in JD.com this March, and I worked as the inaugural president of JD Explore Academy.
You know that JD is a very large corporate group, and it contains a lot of directions.
So we have retail, we have logistics, we have JD technology, and we have JD Health, many others.
Maybe a lot of people can read the news articles to have a good understanding of JD's business.
So it contains so many businesses. But we have to understand all these businesses are actually driven by high technologies.
And you know that JD has a bigger aim and would like to build a more successful business.
So it's important to develop a technological strategy.
It's especially for its future. So this is a reason why they invited me to join the company, and to commence something about fundamental research in artificial intelligence, or maybe digital technologies.
Right. So before joining JD, I was a full professor at the University of Sydney, and I received a lot of recognitions in artificial intelligence or computer science.
And given this background, so I think it's helped me to develop this new academy in JD.
So basically, it's under JD technology, but this academy, I mean, the JD Explore Academy will serve for all JD's business.
And we initiated three main strategies.
They are trustworthy artificial intelligence, super deep model, and also quantum machine learning.
So why we have all trustworthy artificial intelligence, because, you know, so currently, people care about their privacy, and people really want to understand when deep learning works, and when it fails, and why it works, and why it fails.
So there are a lot of situations we want to have good explanations to the working mechanisms of deep learning.
And also, we hope that our deep learning models, all the AI systems can deal with adversarial attacks, can deal with scenario noises.
So there are many situations that make our systems not so stable.
So we need to find ways to stabilize our systems.
And also, you know, that different systems may treat different people, I mean, people from different groups in different ways.
It can be unfair.
For example, some services, some services may treat a female and male in different ways.
This can be unfair to either male or female. And then maybe treats, treats adults, elders, and maybe kids in different ways, maybe also not so, so fair.
But we also need to protect the people from different groups. So given this all situations, we also needed to develop a fair artificial intelligence systems.
So we needed to conduct trustworthy research for JDS technology. And also, you know, that's the GDP, GPT-3 is a very successful.
So a lot of bigger companies, they have their supermodel strategies.
So we know that recently, Google, Microsoft, Alibaba, Tencent, they also have developed super deep models.
So, so JD also think that it's important to have our own supermodels.
Given those supermodels, we can better serve our customers, and we can distill better quality small models.
So that can be deployed to IoT services. So there are many situations.
So we needed to develop supermodels. And also, you know, that's a, the success of the model is due to is intensive computation.
So you know, that's a, but the energy cost is very high.
So in the future, so what is the best way to deal with the energy consumption, that is the quantum computing.
So given this, so in the future, we believe that, so more is different.
So we believe that more is different.
So given this, so we needed to consider how to really explore the quantum computers to deal with the exponential growth of deep learning models.
So this is a reason why we have quantum machine learning strategy.
So we have the three strategies, but they're actually integrated with each other.
And we believe that such kind of integration will transform JD's future technology.
So thank you.
There we go. There we go. With that, Jane, you know, so very, very relevant history you've got and pedigree to be leading this group.
And just like you started with JD.com in March, I started with Cloudflare in March.
And prior to that, I worked at Amazon Web Services running their enterprise business across Asia Pacific and Japan.
And I think about some of the stories you talk about JD.com. And the reason why you are now building this big cloud business is very similar to why AWS was created to help Amazon.
One of the areas I'd want to touch on is, I'm assuming, in addition to some of the innovation and some of the machine learning capabilities and some of the quantitative computing advancements you're building, you're also helping deliver some operational efficiencies and retiring sort of technical depth within JD.com as part of moving that to the cloud.
Is that a strategy that you're working on as part of the bigger organization?
I believe that this will do in the near future.
And to consider how to integrate artificial intelligence and the JD cloud, I think it's better to understand the current advantages of JD cloud.
I'd like to mention some points about JD cloud's advantages.
So specifically, JD cloud has embraced the concept of cloud-native computing since its inception and began collaborating with the Cloud Native Computing Foundation early on, achieving viral containerization.
So JD currently has the world's largest Docker cluster and the Kubernetes cluster and has developed an industry-first hybrid cloud operating system.
We call it JDoS or JD operating system.
JD cloud has embraced the cloud -native computing since its inception. JD experiences massive data and traffic growth during the annual shopping events like 618 and W11.
The largest surge in network traffic brings about unprecedented challenges to both front-end services, including the website, order system, invoicing system, payment system, searching system, recommendation system, and the back-end services, including warehousing, distribution, customer service, and after sales.
Therefore, JD must rely on a flexible, elastic, and scalable platform, which was the determining factor behind JD's early adoption of cloud-native computing.
JD cloud launched JDoS in 2014 to promote its business containerization, gradually connecting businesses to containers and differentiating businesses through containers.
In 2016, all JD services were connected to containers, and the number of container instances reached about 100,000.
By 2018, JD has built the world's largest Kubernetes cluster, upon which it built one of the world's most complex distributed databases, the Vitis cluster.
The JD's container numbers are reaching hundreds of thousands to date.
JD cloud has the world's largest Docker cluster and Kubernetes cluster, and is currently one of the most widely containerized platforms in the world.
From enabling Docker to fully embracing cloud-nativeness, JD cloud's many years of technical practice has not only consolidated its technical foundation, retained more and more technical capabilities, but also safeguarded JD's annual sales events, becoming the technical cornerstone of both events.
Based on the cloud-native architecture, JD cloud operates one of the world's largest container clusters, with an online management support count of over 2 million, and a peak of operating container call count of over 10 million.
Unitized costs are reduced by over 30%, and delivery efficiency is increased by 1 .5 times through economies of scale.
Consequently, the lower costs and the increased efficiency enabled JD cloud to meet more business needs in a limited time frame.
This demonstrates the effectiveness and revolutionary benefits of cloud-native computing, paving the way to digital transformation across the industry.
Building upon JD's own large-scale implementation of the world's most complex cloud-native scenarios in conjunction with innovative exploration in the containerization and cloud-native fields over the past seven years, and the successful adoption of cloud services by multiple enterprise clients, JD cloud released the industry's first hybrid cloud operating system, RINGEN, or Cloud Sheep.
RINGEN uses an enterprise cloud-native container kernel to achieve powerful management and scheduling capabilities.
The Kubernetes-enhanced engine is capable of multi-cluster lifecycle management, mammoth-scale cluster management, unified cloud-native management and operation, and can carry out comprehensive SIEM security management in conjunction with comprehensive link and process security monitoring.
RINGEN provides a consistent multi-cloud container operating environment, multiple path capabilities, and application development platforms, and promotes a threefold increase in the average data center CPU utilization rate, reducing technical investments while greatly increasing productivity.
The JD cloud will continue to increase its investment in cloud-native technology, relying on years of experience with cloud -native technology in practice and internally developed cloud-native service solutions.
JD cloud is capable of enabling the deep integration of cloud-native computing and enterprise management, empowering enterprises through cloud -native technology to help them in digital transformation and the subsequent creation of high commercial value.
We have to understand that containerization is very important for JD to provide high-quality services to our customers and also through the AI strategy so we can provide additional value to our customers.
And eventually we have a good integration between JD cloud and JD cloud's equal customers.
So given this, JD can grow its market quickly. So I think you know that we have a very good collaboration between JD or JD cloud with Cloudflare.
So maybe I guess you want to mention something or introduce something about the collaboration.
Yeah, absolutely. I'm happy to talk about that. I mean, first and foremost, it's great to sort of hear the evolution of JD cloud.
I personally, we see customers ask for modernization of enterprise applications.
We ask for, we see customers ask for leading cloud providers who can think around industry verticals like you're doing.
We ask, you know, cloud providers are asking around containerization and machine learning around putting smarts at the edge and solving real business problems for their customers.
So I think that's, you know, the fact that the JD cloud is moving down that path is really beneficial and us at cloud for a really appreciative of the partnership we have today, but really, I suppose, confident boy around what the future holds.
Back to your question, you know, mainland China and greater China and global partnership with JD is super, super important to Cloudflare.
Now we see the three facets. It's obviously the China commercial market.
It's how we help customers go from China to global.
That's how we help customers come from global into China. And JD's cornerstone partnership to do that at Cloudflare, obviously where there's the technology side of it.
And there's also building out the infrastructure side of it.
So we can have the most reliable, fast, scalable and secure network to serve our joint customers.
So we're super excited about the partnership. It's only just started.
And maybe six months into this business and supporting my greater China team and working with your team.
Again, the opportunity moving forward is significant.
So back to the fact that I'm supposed to be interviewing you, Dacheng, not you interviewing me.
So just in the interest of time. I mean, obviously, you've got a big team, you know, you're very aware of some of the industry trends.
If you think about five, seven, 10 years out, what does good look like for JD cloud?
And where will you be prioritizing if you think about your customers and how and how would you think someone like a Cloudflare can help JD.com and JD cloud maximize the opportunity ahead?
Yes, Jonathan. Yeah. So we're here today to discuss the development trend over the next 10 years.
The current rise of the digital economy shows that technology is the driving force of the industrial transformation.
According for this, the development of JD cloud considers technologies that are currently more closely integrated with the industry.
These technologies can be represented by ABCDE.
Where A means artificial intelligence, B means blockchain and big data, C means cloud computing, D means device or the Internet of Things, and E means exploration.
Exploration to the cutting edge technology.
These technologies are all symbiotic with each other.
So we cannot separate them independently.
If we picture technology as a person, AI constitutes the brain, IoT is a neural terminal, the cloud is the body's trunk, muscles and blood vessels, big data and blockchains are its blood and oxygen, and exploration is analogous to our human's curiosity.
The whole is greater than the sum of its parts when combined together.
These elements form a more capable and competitive organism to evaluate and discover more possibilities in the industry.
So this is the reason why JD's research or technological strategy is the integration or the seamless integration of ABCDE.
So we don't treat them independently, but integrate them seamlessly.
Here, I'd like to take JD cloud's privacy computing technology for example.
So this technology has been widely used in the process of business implementation for transaction data and user identity privacy protection.
No doubt, the privacy computing technology is a part of JD's ABCDE strategy.
So the fusion of privacy computing and trustworthy artificial intelligence, or basically PAN-AI, can achieve simultaneous data availability and anonymity and enhance the value of data security integration.
The cryptography-based intelligence security and privacy protection can achieve secure computing of multi-party data through collaboration of software and hardware.
The cloud-assisted privacy computing that utilizes the computing power, storage, and data resources of the public cloud can practically achieve the secure computing of massive data in the cloud.
Combining technologies such as automated machine learning to reduce technological barriers to privacy computing and truly realize and improve large-scale commercial applications for privacy computing.
Thank you for your partnership.
Thank you for being part of Birthday Week.