Perspective of the 2019 cloud computing battle

On March 21, 2019, the Alibaba Cloud Beijing Summit came to an end. Zhang Jianfeng, President of Alibaba Cloud Intelligence, explained the "four-stage rocket" of Alibaba Cloud's strategic acceleration: the cloud blessed by Bodhidharma, the cloud of data intelligence, the cloud of best practices and the The integrated cloud has caused huge repercussions in the industry. As the producer of "Enterprise Infrastructure Session - CIO Core Capabilities: Building a Billion-Scale IT Infrastructure on the Cloud", I had the honor to preside over the session and gave a 10-minute opening speech. I am very happy to combine the organization of the conference In the process, I will share with you my thoughts on the development of cloud computing in 2019.

Enterprise-level Infrastructure Session - CIO Core Competencies: Building a Billion-Scale IT Infrastructure on the Cloud

Thanks to Xu Xijing's trust and trust, I, together with the special architect Zhang Qiang and another producer Meng Zhen, managed the selection and determination of the topics.

At present, a large part of domestic public cloud revenue still comes from IaaS infrastructure. First of all, this shows that China's cloud computing development stage is still in the early stage, and there is still a lot of room for the recognition of values ​​such as elasticity and efficiency.

There is a saying in the industry: the things that make money today are not cutting-edge, and the things that are cutting-edge at the moment are not making money. And the technology and products of enterprise infrastructure need to make money and be cutting-edge, so how should the theme be arranged?

It should be said that the enthusiasm of the students to release products and introduce technology is still very high. After the initial review, after topic integration, we quickly established 8 speeches.

 "Innovation and Practice of Intelligent Operation and Maintenance of Feitian Infrastructure" by He Cheng

 "Analysis of Enterprise Application Practice Analysis of Building Stable, High Business Elasticity and Low Cost Based on ECS" by Hu Xiaobo

 Alex Chen's "Inclusive Data Smart Economy Next Generation Cloud Storage Innovation"

 "Discussion on Enterprise Hybrid Cloud Storage Architecture in DT Era" by Mr. Wang Ximing, Director of NetApp

 "Building a secure, intelligent and open next-generation enterprise-level network" by Liquan Tan

 Pan Yue's "Heterogeneous Computing Provides Nuclear Power for Enterprises Going to the Cloud"

 Zhang Zixuan and Zhou Jian's "Realizing Large-scale and Automated Cloud IT Management"

 Zhang Jun's "Cloud escort service to ensure peak hours on the cloud"

The main line is very clear, that is to show the billion-level capabilities of Alibaba Cloud's technical facilities from the dimensions of IDC, elastic computing, storage, storage ecology, network, GPU heterogeneity, OpenAPI, cloud escort, etc. The logic is to display our capabilities layer by layer from bottom to top The latest technology and products to enhance customer interest! Special thanks to Mr. Wang Ximing for bringing NetApp's thoughts on hybrid cloud storage, I also benefited a lot.

why do you want me to start

My original intention was to find a beautiful classmate to open and host the show. Xu Xijing pointed out that I am well-known in the industry. If you invite me to start the speech, it will attract more target groups to sign up for the conference; at the same time, Xu Xijing greatly appreciates the hosting style of the industry media celebrity "Bao Ge", that is, after each speech Afterwards, the host will make a brief comment to help the audience understand, and I hope I can follow along.

As a producer, of course, the value of the special session should be regarded as the first value, so I actively accepted the task of opening and hosting.

So, I think that since we are going to start, there are two things to make clear. The first is the arrangement logic of this special speech.

Generally speaking, there are ten trends, no more or less. So I patted my head and thought about 14 items, and posted them in the internal nail group to collect feedback from students in the department. As a result, the following ten items were settled, hoping to resonate with everyone.

Around 2020, China will have a break-even public cloud

In my opinion, the break-even point of the public cloud is critical to the long-term growth of the company. On July 24, 2015, the global e-commerce giant Amazon released its second quarter financial report for 2015. Amazon, which has been famous for its losses, finally made a profit. AWS contributed $1.824 billion in net revenue in the quarter, an increase of 81% year-over-year; AWS operating profit also increased to $391 million from $77 million a year earlier.

Shares of Amazon jumped nearly 20% in after-hours trading after the earnings report. After that, AWS's revenue and profitability both rode the red dust and grew rapidly. It was also during this time period that Alibaba Cloud's "liar" became a "god". Soon after, CCTV broadcast a feature film "Dr. Wang's Computational Dream" on March 7, 2016 in the column "Approaching Science".

Note the multiple key words in this assertion:

 Around 2020. There will be no drama in 2019, and the reason can be seen from the financial reports of various public cloud companies. Our internal classmates told me: After ten years of hard work, we have controlled the cost to the extreme, we are not profitable, how can others be profitable.

 China. As mentioned above, AWS has already become the first profitable company in the world.

 Public cloud. In fact, it is difficult to say that it is difficult to be a profitable cloud company in China, and it is simple and easy to say. I know that some small companies, with a small number of people, receive a few small orders a year, and they are profitable on the books. But the problem is that these orders are all through previous relationships and do not have the potential to expand. Therefore, a temporary break-even is meaningless, and a long-term quick profit from the break-even is the key.

In the IT industry, especially the winner-take-all field, there is a phenomenon that has been proven many times. The first place is profitable, the second place is not profitable, and the third place and after all lose money. At the end of 2018, several cloud computing companies in China fought a war of words, saying that they were the Nth place, which is completely meaningless PR behavior.

Another method is to use financial methods to adjust income. The statement may show that the financial statements in recent years are all slightly profitable. Without seeing a detailed report, it can be guessed that a large part of the revenue comes from private cloud orders or hardware sales, covering public cloud losses in the short term.

What I mean by "break-even public cloud" refers to truly making profits from IaaS, PaaS and other fields, so as to have long-term development value, rather than a short-term catering to investors. I insist on recognizing the Matthew effect in the public cloud field, that is, the strong become stronger and the larger become stronger.

If the development goes smoothly, the time node for Chinese public cloud vendors to break even will be 2020. If it is slower, it will most likely be 2021. Only when there is an unstoppable advantage will companies withdraw from this field. In the United States, many companies have withdrawn from the public cloud field in the past few years.

Managed hybrid cloud through open APIs will be more popular with large customers

The development trend item of this article has been prepared a long time ago, and I have not participated in the preparation of the keynote speech materials of President Zhang Jianfeng, but I am happy to find that the speech at the main venue in the morning has paved the way for this item. Because this article and being "integrated" as a first stage rocket are complementary. "Being integrated" is a company strategy, so how is it integrated tactically? The means is through OpenAPI.

A few days ago, a classmate from the Technology Strategy Department shared with me a photo of a speech by Dr. Song Jiqiang, Dean of Intel China Research Institute. The "Unified Heterogeneous Computing Software Development Interface" OneAPI and the OpenAPI mentioned here are also in the same line. .

In mid-2018, the development of Alibaba Cloud OpenAPI began to accelerate. This shows that our "integrated" is not proposed suddenly, but after a long period of internal deliberation.

Why have OpenAPI? Because Alibaba Cloud provides resources first, including computing, storage, network, etc., we also provide an official background control and management interface. To control one machine, ten machines are simple, but it is more troublesome to control hundreds of foreground machines through the official background.

At the same time, the major customers of GC6 and GC7 often have strong technical R&D teams and are very clear about their own business. They prefer to develop their own console and flexibly call various resources provided by Alibaba Cloud. Without OpenAPI, this kind of development would not be possible.

Back then, Taobao’s Jushita also provided an API for merchants to customize their own backend, thus achieving great success. Once large customers feel the convenience brought by API, they cannot leave.

Further, AWS has been committed to providing excellent, unified, and efficient APIs to help large customers make better use of resources. When many foreign customers and partners enter China, they assume by default that Alibaba Cloud will have a systematic OpenAPI that can be called.

I once said that the public cloud is the end, but the hybrid cloud is an inevitable historical stage, and it is now in this historical stage, so to meet the needs of the hybrid cloud, we must also provide OpenAPI to better realize Alibaba Cloud's "integrated" ".

Customers tend to use cloud storage disaster recovery technology to ensure data reliability

Following the hybrid cloud mentioned in the previous article, users are very interested in hybrid cloud, mainly because of elasticity. Elasticity is a fundamental characteristic of the cloud, and a cloud without elasticity is not really a cloud. The way to achieve elasticity in a private cloud is to build a hybrid cloud.

At the same time, hybrid cloud can also bring high reliability of data. In 2018, there have been many disasters in the Chinese market due to unstable storage, and in some extreme cases, data is completely lost and cannot be recovered. After the incident, customers' demand for data reliability and their willingness to invest has greatly increased.

The reliability of Alibaba Cloud storage is 11 9s, and in the 3AZ environment, the reliability reaches 12 9s. What should I do to ensure that the user's data is never lost? I don't think that asking customers to buy a mobile hard disk to back up data regularly will be the future development direction of cloud computing. I think that customers tend to use cloud storage disaster recovery technology to ensure data reliability.

Director Wang Ximing of NetApp introduced the difference between backup and disaster recovery. Backup is a term with a long history, which is to make a copy of data on another medium. If the original data is lost, just copy it back. Disaster recovery is a term of cloud computing. Disaster recovery refers to preparing a complete copy of your private cloud data and code on the cloud. Once a problem occurs in the local private cloud, whether it is data loss or virtual machine crash, The public cloud can immediately pull up the computing environment and take over the service to the customer, so that the end user does not feel that there is a problem with the private cloud. Therefore, the storage here is not only a backup of data, but also a backup of the entire computing environment and the ability to pull up in real time.

Alibaba Cloud's hybrid cloud storage team has developed the HDR (Hybrid Disaster Recovery) solution after two years of hard work. The specific technologies will not be repeated one by one. I will emphasize here that the HDR solution is not only a disaster recovery solution, but also can solve the problem of customers' relocation. Simply put, using half of the functions of HDR, the customer's private cloud can be migrated to the public cloud; then using the other half of the functions, the customer's private cloud and Alibaba Cloud public cloud constitute a disaster recovery solution. Migration to the cloud is only a subset of disaster recovery technologies.

China's first large-scale company that is truly fully cloud-based

What is comprehensive cloudification? I think there are two key indicators: First, the company's core business must be placed on the cloud. Many companies, especially financial and telecommunications companies, use virtualization and cloud services in the peripheral and middle layers. Native and other cloud computing technologies, but the core trading system is still running on traditional mainframes, which is not a full cloudification; second, there can only be one cloud within the company, which means that all resources are scheduled and managed in a unified manner. Duoyun is not really comprehensive cloudification.

Every successful Internet company has its own technological development process. In the process of rapid business development, it is how to achieve it effectively and how quickly. As a result, the architectural design of the infrastructure is not very elegant, or it does not meet the requirements of modern cloud computing operating systems. Migrating the original business to the cloud requires time and detailed analysis and planning without affecting the existing business. Blacksmithing needs to be hard on its own. Public cloud companies will be most convincing if they put all their business on the cloud.

As President Zhang Jianfeng pointed out, 60-70% of Alibaba Group's business now uses Alibaba Cloud's resources, and in the next 1-2 years, 100% of Alibaba Cloud's resources will be used. If successful, this will be a historic watershed. Because as far as I know, many cloud computing companies have multiple completely isolated clouds inside, and the integration is still far away, and there is no schedule. Comprehensive cloudification is easier said than done.

Enterprises with sharply reduced budgets for self-construction have become the main force in the public cloud

I've written a few articles about how cloud computing can help businesses save money, and I've been seeing this trend come to fruition lately. In order to complete the digital transformation and enjoy the technology and cost dividends brought by the digital economy, it is the first driving force to urge enterprises to go to the cloud.

Cloud computing is no longer a question of whether it is on or off, but a question of morning or night. If you do not keep up with the pace of the cloud era, you will be abandoned by the era.

The economic situation in 2019 is severe, and many companies have begun to consider putting offline computing clusters and self-built cloud computing on the public cloud. Many customers stopped self-build because of the high cost. While slowly digesting the original self-built machines, the rest is the question of how much business to put on the public cloud.

Before 2019, these customers preferred to own cloud assets themselves. The willingness to consider going to the cloud indicates that the customer has encountered a problem, which is largely due to the substantial reduction of the self-built budget. Even companies with good technical capabilities will choose the public cloud for financial budget reasons, and everyone should actively prepare for it.

Looking back at the history of cloud computing in China, the period of explosive growth of public cloud has not passed, and there are still a large number of enterprises whose traditional IT has not been on the cloud. Now many companies will directly use the public cloud directly beyond the stage of self-build.

Kubernetes and Kata Containers will become mainstream in the industry

Kubernetes is a container management and orchestration tool, and it has been widely recognized as a representative project of cloud native. Alibaba Cloud is a platinum member of the CNCF Foundation, the owner of Kubernetes.

Today we will talk about Kata Containers. There is a recent news that the founding team of Hyper, Zhao Peng, Wang Xu and other students joined Ant Financial. Hyper is also the main contributor of Kata Container. However, what we talked about today about awareness and use of Kata Containers is not directly related to this news.

In 2015, Intel announced the Clear Containers open source project, which was merged with Hyper's runV at the end of 2017. The new project, called Kata Containers, is fully compatible with the Open Container Initiative (OCI) specification and Kubernetes' Container Runtime Interface (CRI). By the way, Kat released in January 2019 a Containers version 1.5 already supports the Firecracker hypervisor released by AWS.

What are Kata Containers? Four words: safe container. Docker containers are great, but they don't solve security problems. The solution to the security problem is to take advantage of the virtualization capabilities of the CPU hardware. Kata Containers combines the agility of containers with the security of VT hardware virtualization.

Scale effects of public cloud bring irrefutable cost advantages

I have seen many experts explain that private cloud is cheaper than public cloud from various dimensions, so I will not analyze them one by one here. In fact, I am very puzzled, because this logic does not make sense, because the amortization of costs caused by the large-scale use of cloud computing is one of the important means for cloud computing to improve IT efficiency. If scale does not bring the price advantage of higher-level SLAs, it fundamentally negates the value of cloud computing.

First of all, these experts often do not calculate the personnel, maintenance and other expenses brought by the self-built private cloud. The use of public cloud is a package payment, after payment, electricity, space, computer room and other expenses are included.

In addition, the flexibility of computation is also an issue. A customer who needs an average of 10 physical machines per day for computing will not purchase 100 physical machines just because one day out of 365 days of the year will require 100 machines. It is impossible to enjoy the peak computing power of 100 machines without buying a private cloud with one more machine. To use the cloud, you have to pay for the elasticity and efficiency that the cloud brings.

The open source ecosystem has become a means for cloud vendors to maintain developer relationships

I have been in contact with Linux since 1995 and have been active in the open source community for a long time, and have deeply experienced the flesh-and-blood connection between open source and developer relations. I have also said many times that independent research and development and open source are not contradictory, which means that cloud manufacturers can better use the open source ecosystem as a killer to strengthen the connection with developers.

Former Microsoft CEO Ballmer once shouted the slogan "Developer, Developer, Developer" at the Microsoft Conference, and he recognized that getting developer support is the key to Microsoft's success. The success of a cloud platform is also its ability to gain broad developer support.

The China Open Source Hackathon (once every six months), which was initiated and founded in April 2015 while I was at Intel, has persisted until the ninth session. This kind of activity will become mainstream in China. I am more inclined to join the technology alliance, and speaking in code is the basic communication method of the open source ecosystem. The China Linux Kernel Developers Conference, which I took over in 2007 and started in 2006, will come to Hangzhou for the first time in 2019. I hope to bring you lectures and exchange opportunities from top Linux kernel masters.

At present, the developer relations of many companies are still slogans and have not been implemented. Since the birth of open source, many excellent open source projects have emerged, and these projects have become the de facto standards in the industry. Helping more open source projects succeed on cloud platforms is an important means of leveraging developers. An important indicator of success is whether it can bring real economic benefits. Developer relations and open source ecosystems not only bring fame to developers, but also bring real profits to developers. An ecosystem in which only the cloud platform earns its own money and other participants only earn experience and no money is an unhealthy and unsustainable ecosystem.

The breadth and depth of cloud GPU computing applications in business will become a new yardstick for measuring enterprise competitiveness in the next decade

On December 25, 2018, Stanford University released the latest DAWNbench deep learning reasoning list. Alibaba Cloud won the double championship in image recognition performance and cost, breaking the 8-month record based on the Amazon AWS computing platform. This is the first time a Chinese technology company has appeared on the list.

Yes, this is the first time a Chinese company has entered the list, and I believe it will not be the last. This just opened the curtain of cloud GPU computing's strong landing in the cloud computing market.

Computing power not only refers to the computing power of the CPU, but also includes the computing power of the GPU, especially in the public cloud.

With the rise of artificial intelligence and entering the stage of large-scale practical use, the computing power of GPU will become the basic indicator of enterprise competition. How to better use the cloud GPU to quickly and economically meet the computing needs of the business is a question worth considering.

The key technical capability of the city brain is to process the massive videos generated by high-definition cameras at high speed, and the GPU capability that can process thousands or even tens of thousands of videos will be very necessary. As a modern technology enterprise, especially an artificial intelligence-related enterprise, it is better sooner rather than later to make good use of cloud GPUs.

Many small and medium-sized cloud manufacturers have registered on the Science and Technology Innovation Board, but long-term development depends on technology

I believe that many of my friends have opened accounts on the Science and Technology Innovation Board, which is actually to open the trading authority of the Science and Technology Innovation Board on the Shanghai Stock Exchange account. As the name implies, the Science and Technology Innovation Board is a stock trading board established to encourage technological innovation. The rules are very different from those of the Main Board, the Small and Medium-sized Board, the Science and Technology Innovation Board, and the New Third Board. Here I will talk about the impact of the differentiated IPO conditions of the Sci-Tech Innovation Board on cloud manufacturers. Simply come, from the general direction guidance, financial profitability indicators, etc., the Sci-Tech Innovation Board will remove all obstacles that hinder the IPO of small and medium-sized cloud manufacturers. .

On January 30, 2019, the "Science and Technology Innovation Board IPO Registration Management Measures (Trial)" issued by the China Securities Regulatory Commission pointed out: "The Science and Technology Innovation Board focuses on supporting the new generation of information technology, high-end equipment, new materials, new energy, energy conservation and environmental protection, and Biomedicine and other high-tech industries and strategic emerging industries promote the deep integration of the Internet, big data, cloud computing, artificial intelligence and manufacturing."
This is to directly encourage cloud computing companies to apply for IPO listing.
The "Shanghai Stock Exchange's Science and Technology Innovation Board Listing Rules" issued by the Shanghai Stock Exchange on March 1, 2019 stated:
“2.1.2 When an issuer applies for listing on the Sci-tech Innovation Board of the Exchange, its market value and financial indicators shall meet at least one of the following criteria:
(1) The estimated market value is not less than RMB 1 billion, the net profit in the last two years is positive and the accumulated net profit is not less than RMB 50 million, or the estimated market value is not less than RMB 1 billion, and the net profit in the most recent year is Positive and operating income not less than RMB 100 million;

(2) The estimated market value is not less than RMB 1.5 billion, the operating income in the last year is not less than RMB 200 million, and the cumulative R&D investment in the last three years accounts for not less than 15% of the cumulative operating income in the last three years;

(3) The estimated market value is not less than RMB 2 billion, the operating income in the most recent year is not less than RMB 300 million, and the cumulative net cash flow from operating activities in the last three years is not less than RMB 100 million;

(4) The estimated market value is not less than RMB 3 billion, and the operating income in the most recent year is not less than RMB 300 million;

(5) The estimated market value is not less than RMB 4 billion, the main business or products need to be approved by the relevant state departments, the market space is large, and phased results have been achieved so far. Enterprises in the pharmaceutical industry need to have at least one core product approved for Phase II clinical trials, and other companies that meet the positioning of the Science and Technology Innovation Board must have obvious technical advantages and meet corresponding conditions. "

Note that profit requirements have been removed from 2 to 5. This is compared to the Nasdaq in the United States, where a new generation of IT giants such as Microsoft and Intel was born, and at that time, these companies could not be required to make profits.

It can be seen that the fourth and fifth items are tailor-made for domestic medium-sized cloud computing manufacturers.

3. Local governments’ financial incentives for IPO companies on the Sci-Tech Innovation Board

To encourage local companies to list on the STAR Market, many provinces and cities have introduced a number of subsidy policies. The highest subsidy reaches 20 million yuan, and the provinces and cities with the lowest subsidy also have 1-2 million yuan. The Xuhui District of Shanghai can give subsidy to the construction unit of the landing project not exceeding 50% of the total project investment, and the maximum annual subsidy is not more than 20 million yuan.

Subsidies of real money will greatly increase the enthusiasm of cloud computing companies to go public.

4. On March 1, 2019, the Shanghai Stock Exchange issued the "Science and Technology Innovation Board Stock Listing Rules of the Shanghai Stock Exchange" stating:

"2.4.5 If the core technical personnel of a listed company reduce their pre-IPO shares, they shall abide by the following provisions:

(1) The company's pre-IPO shares shall not be transferred within 12 months from the date of listing of the company's stock and within 6 months after resignation;

(2) Within 4 years from the expiration of the lock-up period of the pre-IPO shares held, the pre-IPO shares transferred each year shall not exceed 25% of the total pre-IPO shares held by the company at the time of listing, and the reduction ratio can be used cumulatively;

(3) Other provisions of laws and regulations, these Rules and the business rules of the Exchange on the transfer of shares of core technicians. "

Note: The lock-up period for core technicians' shares has been adjusted from 3 years to 1 year. After the expiration of the period, 25% of the pre-IPO shares can be reduced every year. This means that the core personnel of the cloud computing company can start to sell part of the company's stock to cash out after one year after the listing on the Science and Technology Innovation Board. This is also a strong encouragement for cloud computing companies to go public. After all, employees can sell stocks regardless of whether they are profitable or not.

My judgment is that in 2019, there will be more than one small and medium-sized cloud computing manufacturer registered on the Science and Technology Innovation Board to obtain financial support for continued development. But what I want to emphasize here is that the competition in cloud computing will be a protracted war, a war that relies on technology to confront hard. This is not only responsible for customers, responsible for the company itself, but also responsible for the healthy development of the entire Chinese cloud computing industry.

From another point of view, the Science and Technology Innovation Board adopts a registration system. At the same time, there is no increase or decrease in the first five trading days, and the daily increase or decrease is 20%. This means that while financial transparency, investors will have greater voting rights, which can also greatly reduce the impulse to "cut a handful of leeks and leave".