The history of cloud computing development of Internet giants has been written a thousand times.

However, the opening of the official history is always artificially decorated, and it is more than wonderful, but it only retains the overall rationality, and does not reveal its partial unreasonableness.

The birth of the cloud, naturally, is not entirely the result of “business growth, forcing technology upgrades” in the official context.

After all, decisions are made by people, and people cannot always be rational.

The moment when it really plays a decisive role, the moment when the decision is made, the decision makers of the enterprise need to be hot enough, with sensibility, impulsiveness and unreasonableness, to sign these few words that change the direction of the company’s strategic development.

The prequel of cloud computing is composed of stories behind this sensibility, either ups and downs, unintentional, or unreasonable stories.

Taobao and Alibaba Cloud where the network card is smoking

Talking about the history of cloud computing, most of them start from Ali. It is well known that, between 2006 and 2009, Amazon’s AWS was in its infancy, Ma Yun and Zeng Ming discussed cloud computing, Wang Jian joined, and Alibaba Cloud entered the stage of history.

However, this step of cloud computing is not all Alibaba’s initiative and planning in advance.

Back in 2006-2008, Taobao Mall (predecessor of Tmall), which was established shortly after the project, was still in a “stocking mode” within the company – no resources, no one to care, employees only felt that the department was not very popular , I can’t imagine what kind of “hell mode” I will experience in 2008.

In fact, in 2007, Taobao’s business suddenly began to explode. By 2008, it had grown at the level of nuclear explosion, and the old IT system had been forced to the brink of collapse. It is said that between 8:00 and 9:30 in the morning, the processor utilization rate of the server often soared to 98%, which was two percentage points away from bursting.

During this period of time, Taobao, which was “covered with injuries”, allowed a group of top technical talents to grow rapidly, such as Ali programmer Daniel Lin Hao (Bi Xuan) and others.

A number of Taobao Mall technical veterans of the same period vividly described those “tragic” pictures to Leifeng.com:

“Taobao was a pioneering business form at that time, and the technical components, performance and capacity on the market were not worthy of Taobao. The most terrifying time was the big promotion. When the traffic peak came, all the machines in the computer room went down, and the Gigabit The old network card ‘popped’ and smoked. Immediately replaced it with a new batch of network cards, ‘popped’, smoked again and died.”

Why was Taobao so “crisp” at that time? In the old architecture, all the logic and processes were very simple: as soon as the business requirements came, a system was supported online, and no architecture, platform, or redundancy was considered. It was only when the new architecture emerged that the platform as a whole was revealed, and the well-known concept of China-Taiwan was gradually settled at this stage. The new-generation system didn’t really stabilize until 2013, and it has been “tinkered” all the way to this day.

Taobao is the one with the fastest development and the most problems, but it is not only Taobao that has problems. Alipay, which undertakes the transaction settlement function, and Ali’s internal financial system also suffer from “suffering”. “In the early days, the accounts here were not settled a few times,” recalls a former Taobao employee.

In addition, it was also in the stage of evolution of Taobao’s technical architecture system 2.0 to 3.0. It can be said that the entire technical level of Alibaba was under huge pressure, which was called the biggest technical “life and death disaster” since Alibaba was born.

“Every day from opening to closing, I just keep revising and fighting fires. I have just sorted out my work, but the situation has changed again, and everything has to be sorted out again. Getting off work at one or two o’clock in the morning is the norm. The class is over.” The technical veteran bluntly said that this is the most fulfilling and painful memory in his career. Halfway through the project, many colleagues are almost depressed, and they feel like they are going through a shaky life. the bridge, the cargo on the back is getting heavier and heavier every day.

It is no exaggeration to say that the emergence of this round of “calamity” is the real beginning of Alibaba’s entry into cloud computing.

Since then, it has been the vigorous process of Alibaba’s “going to IOE”, the cloud computing system “Flying” was born, and the “Cloud Ladder Project” was successfully realized with the efforts of the whole group. A series of inspirational stories were staged on Alibaba Cloud.

Wang Jian

In 2009, Alibaba Cloud successfully passed the major version upgrade, and the performance of Feitian system was acceptable. However, between 2010 and 2012, Alibaba Cloud stagnated, and the entire department scored the lowest score within the group for three consecutive years. The most well-known scene is when Wang Jian burst into tears at the annual meeting and said, “I have been scolded even more in the past two years than I have in my entire life. However, I do not regret it.”

Beneath this appearance is the immature Alibaba Cloud ‘s concerns about productization and commercialization .

Many people agree that Alibaba Cloud has a group of technical “madmen”, which not only means that they have extremely high enthusiasm and attainments in technology, but also means that they may not be proficient in things other than technology.

A former Tmall technical expert who has dealt with Wang Jian commented that Alibaba Cloud at that time was not technically incapable, but productization and commercialization can even be described in four words, one less and business department ” Empathy” mind.

This group of P9 and P10 cloud technology experts will polish their systems at any time. From a purely technical point of view, it is always right to strive for perfection and to find and solve problems as early as possible – but for the business department, this kind of “polishing at any time” without prior greeting is simply a disaster. .

“For example, they will suddenly go to the computer room and change two parameters to see the effect, but the business side will crash for two hours, which has reached the standard of P0 level accident. If it is in a department with strict management, the equity will be deprived long ago. , Let the experts pack their bags and go home.” The expert said with tears in his eyes.

Further benchmarking with AWS will reveal that Alibaba Cloud’s products, from details to applications to business cooperation, have many loopholes, and the management process is rather rough. Later, Alibaba Cloud spent a year or two working with other departments to polish it. Modifications have formed today’s productization level.

The launch of “Jushita” in 2012 is a typical growth experience of Alibaba Cloud and business departments working hand in hand. To some extent, it can also be regarded as a milestone for Alibaba Cloud to become mature and commercialized.

In the early days, large and medium-sized merchants had their own business process models and IT systems, so they were unwilling to rely on others to sell and manage products in accordance with Taobao’s background requirements. At this time, Taobao needs to be a platform oriented to the big B-end, connecting large and medium-sized merchants to form a synergistic relationship.

However, the first version of this platform was not recognized by merchants, and was evaluated as “very technical, but not commercial.”

In order to grasp the pain points of the merchants, it is said that the team visited all the merchants above the waist in Hangzhou at that time. During this process, the real needs of the merchants surfaced: the business is developing rapidly, but the IT level cannot keep up.

As a result, this set of cloud hosting services for merchants’ IT systems and data was born, which is the later “Jushi Pagoda”.

To put it simply, as long as merchants focus on doing business, they do not need to deal with IT problems in time. A series of problems such as IT resources and system stability, order fulfillment speed, etc., under high traffic, are handled by Jusita.

A lot of cloud resources are needed in the process, and Taobao Mall naturally pulls Alibaba Cloud into this project.

“If there was no help from Taobao Mall/Tmall to polish and implement the scene at that time, Alibaba Cloud may not have the status it is today. ” A former Alibaba veteran said.

According to statistics, on Double Eleven in 2012, nearly 70% of Tmall orders were successfully used on Jushita. This is a project that is highly rated within the group. It has solved many old problems for Ali. For example, the supply chain data of various industries that was difficult to obtain before can finally be deposited on the stone tower. Such a project was later taken over by Wang Jian, which shows the growth performance of Alibaba Cloud.

After successively overcoming the “life and death” of technology and business, Alibaba Cloud, under the leadership of Wang Jian and the assistance of Taobao, has gradually and steadily stepped on the right track, becoming the vanguard of cloud computing in China.

In the midsummer of 2015, the Beijing Summit of the Yunqi Conference opened, and the new team of Alibaba Cloud made its debut. Wang Jian slowly handed over the heavy burden of Alibaba Cloud to Hu Xiaoming, and the most turbulent era of cloud computing in China kicked off.

Liu Qiangdong, who covets Dropbox, and JD Cloud

If Alibaba Cloud was born in the torrent of business, then JD Cloud was born in Liu Qiangdong’s thought.

The Jushi Pagoda mentioned above started in late 2011. Coincidentally, the origin of JD.com and the cloud also began in 2011 .

In the official version, the first year of JD Cloud is 2016. Indeed, this is a high-spirited start: within the year of 2016~2017, the public cloud business was officially launched, and Shen Yuanqing joined smoothly; Liu Qiangdong even delivered a surging speech, shouting “Technology! Technology!” ! Technology!” slogan.

But in fact, according to a former JD.com executive recalled to Leifeng.com, the most direct reason why Liu Qiangdong got the idea of ​​cloud is that in 2011, Liu Qiangdong saw Forbes magazine published a business data in October— – Dropbox, a cloud disk manufacturer founded in 2007, has 50 million users worldwide, and its valuation is very high.

Who is Dropbox? In the rumors, Apple even made a 9-digit acquisition offer for it. Its growth in that year can be called rocket speed: before the spring of 2008, there was not much news about it on the market. It was not until Sequoia’s multiple rounds of betting that it became popular rapidly, and it reached the level within five years of its establishment. The “small target” with more than 100 million users has attracted the envy of countless people.

“We have been doing e-commerce for so many years and have not made money yet. This online disk business is popular and growing fast. Let’s try it out.”

This is where the prototype of JD Cloud came from. “The earliest internal positioning of JD Cloud is the online disk. The related file management system is still in use by JD.com.” The former executive confirmed this to Leifeng.com.

He Gang came to JD.com, and his initial starting point was to be a personal cloud disk. This is the real starting point of JD Cloud. As for some later statements, such as the starting point of JD Cloud is to integrate JD’s IT resources, etc., they are all slightly modified versions.

In 2012, there were not many people in China who could independently host cloud computing. He Gang, as one of the heads of the cloud computing business of Shanda Innovation Institute at that time, was invited by Liu Qiangdong to JD.com. The other two leaders of the project are Ji Xinhua and Xu Shiwei, who founded UCloud and Qiniu Cloud respectively.

He Gang

It is worth mentioning that He Gang, Liu Qiangdong’s prime minister, is not unrelated to his former position as the head of the Amazon cloud computing project. As we all know, Liu Qiangdong never hides his admiration for Amazon. In many JD SEC (Strategic Executive Committee) meetings, Liu Qiangdong will compare what Amazon has, what Ali has, and what we don’t have.

It was not difficult for He Gang to make a cloud disk, but due to the low overall technical reserves of JD.com at that time, the funds for technical investment were very limited, and there were not enough technicians in the market. Various reasons led to the early JD.com The experience of cloud disk is very poor, at least there is a gap with the mainstream products at that time.

The earliest JD cloud disk was even embarrassed to open it to C-end users. In addition to internal use, it was also provided to third-party merchants on the JD platform (internally known as Pop merchants) for trial use for them to store some sales materials such as posters and videos.

However, for a large number of traditional enterprises and merchants at that time, the process of informatization had not yet started, the IT capability was almost zero, and the business management even stayed at the level of using Excel. JD Cloud Disk is open to Pop merchants first, and it is logical to support them to launch smoothly on JD.com.

If the level is not enough, it is common for acquisitions to come together.

It is said that JD.com had the idea of ​​acquiring 360 online disk to strengthen the development of its cloud disk business. However, Zhou Hongyi invested only 2 billion yuan for the 360 ​​network disk, but offered a sky-high price of more than 5 billion yuan, far exceeding Jingdong’s expectations, and the acquisition had to be abandoned.

Another failed intended acquisition project was IDC Enterprise Sinnet. An insider close to the project recalled with regret that at the time of the investment committee discussion, Liao Jianwen, chief strategy officer of JD.com, directly rejected the acquisition.

Of course, cloud disks are just the starting point. When did JD.com consider the real upgrade from “JD Cloud Disk” to “JD Cloud”, and consider cloud computing as the backbone to support JD’s entire technical system – where is this important watershed?

This time point, the earliest may be as far back as 2014 .

In an external presentation in 2014, He Gang gave a very clear idea of ​​Jingdong cloudification. It is not difficult to see that at least at that time, He Gang had completely designed the “three-step policy” of private cloud, public cloud and cloud ecology.

As written in the article “Intercepting the “Cloud” Giants of Alibaba and Tencent: Conspiracy to Merge, Difficult Transformation” , He Gang himself is unwilling to limit himself to the “small” business of cloud disk, but wants to switch the cloud to the mainstream of JD.com within the technical support system.

From another perspective, Liu Qiangdong also tends to put this burden on He Gang’s shoulders.

As a company without a strong technical gene, JD.com has long lacked technical talent and budget. In many cases, in response to the big promotion, the technical team has taken a large number of temporary remedial measures that cannot be undone.

How bad is Jingdong’s IT foundation? There is a little story to share: JD.com tried to go public as early as 2012, and investment banks need them to provide the latest various types of operating data from time to time, and JD.com’s IT system takes at least a week to get the data out. The lagging IT capability has also become one of the reasons for the failure of the listing at that time.

However, despite the fact that it is just necessary to go to the cloud and the cloudification route is clear, He Gang faced a lot of pressure in the few years after starting in the name of “JD Cloud”, and his situation was difficult. We are also in “Intercept” A detailed analysis is given in the article. In general, judging from the details told by many parties, the trust Liu Qiangdong gave to He Gang is far from the burden he gave to He Gang.

“With little investment, the direction changes every day, which is very frustrating. His department has a very low sense of presence, and he is often seen only when he is scolded. In the process of promoting reform, he has suffered a lot of grievances. Sometimes Lao Liu criticized His voice can be heard clearly from outside the door,” a former executive revealed.

Until 2016, the fifth year of He Gang’s entry into JD.com, JD Cloud, which has been bumpy since its inception, was finally able to enter the market competition.

More importantly, after listing, JD.com has more sufficient funds and talents to develop cloud computing; Liu Qiangdong also hopes to use cloud computing business to add more aura of “technology company” to JD.com, so that the market-to-sales ratio will double again. . This is the scene at the beginning of this chapter: the public cloud service was officially launched in April of this year, and JD Cloud finally began to exert its strength.

But an imminent question also came to the table: As a latecomer, can JD Cloud find another way?

To know that the public cloud market is inseparable from the consumption of the “four components of the cloud” (server, storage, bandwidth, CDN), the JD.com technology strategy department at that time also believed that a “main focus” must be found to ensure that JD.com can The “Four Pieces” were successfully output.

The first way given is to take advantage of the “Dongfeng” of live video to grab the cloud market . However, in the field of e-commerce logistics, which JD.com is good at, it faces structured text data, and has limited ability and confidence in data processing of multimedia such as video. And Kingsoft happens to have rich experience in video encoding and decoding, which is one of the reasons why JD Cloud wants to merge with Kingsoft Cloud later. Regarding the details of the merger of Jingdong Cloud, Kingsoft Cloud and Baidu Cloud, Leifeng.com has a detailed report in “Intercepting the “Cloud” Giants of Alibaba and Tencent” .

After the merger of Jingdong Cloud and Kingsoft Cloud was vetoed by Liu Qiangdong, the company found another way to combine cloud and AI , and a large number of fellow AI talents such as Zhou Bowen joined as a result. The mighty “AI iron army” is lined up within the group, and JD Cloud has also ushered in the most star-studded period of technology brands.

“Super Girl” Sequelae and Baidu Cloud

In the eyes of the outside world, Baidu Cloud’s experience is usually described as this: Robin Li talks about “new wine in old bottles”, a story of being late except for missing out on cloud computing.

However, the origins of Baidu and cloud computing, strictly speaking, even earlier than Shanda Innovation Institute and Alibaba Cloud. The beginning of this story may have to start in 2005.

In August 2005, there were two major events. First, Baidu successfully landed on NASDAQ in the United States, which became a landmark event for China’s Internet expedition overseas. Second, the finals of Hunan Satellite TV’s “Super Girl” came to an end, and Li Yuchun won the championship. , that session of Super Girl is still the peak of traffic in the history of Chinese TV.

There is a connection between these two things that has to be mentioned. That year, the traffic of Super Girl overwhelmed the vast majority of domestic communities and forums. Baidu Tieba was the only online discussion area that could withstand it at that time, so it quickly became a fan. main base.

This is completely beyond Baidu’s expectations. All the data in their internal testing is incomparable to Li Yuchun; in order to reduce the burden on the system, the administrator deleted Zhang Liangying, a high-rise building with more than 300,000 posts, which made everyone unhappy. , they were shocked to realize that the leader of the fans who came to negotiate with this was actually a college professor…

The shareholder style of Super Girl is a highlight in the history of the rise of Tieba, which impressed the old Baidu people deeply. At the same time, this also reflects the fact that Baidu has accepted the test of high concurrency earlier than any other cloud giant.

Not only is the post bar hot, but in the two or three years after Baidu’s listing, new products have emerged one after another. Baidu people gradually realized a problem: can’t all the components of each new product start from scratch, can they find some common and reusable basic modules or services to reduce the repetitive work of engineers?

So a small department called ibase was established, with Hou Zhenyu (now the vice president of Baidu Group) as the person in charge. Its function is to unify the technical base of Baidu’s non-search business to the greatest extent and establish some basic general libraries – although there was no The term “technical base”.

Hou Zhenyu once told Leifeng.com that from the perspective of building a universal technology base, the history of Baidu Cloud should be traced back to 2006.

In 2006, Baidu World Conference released Baidu’s blog product – Baidu hi. Baidu hi began to provide users with image storage and display functions, among which the image storage function was made into a universal storage module.

This is the origin of Baidu Cloud, and it is also the prototype of the earliest C-end-oriented cloud storage system in China’s Internet.

“Baidu Netdisk’s current large storage also started from the so-called universal storage module on our cloud back then.” Hou Zhenyu said: “Modularization of some basic functions has made our research and development more efficient, and it is from then on. , A new department was established in 2008, which can be said to be an upgrade of the previous ibase small department, to focus on how to provide specialized general services and bases for Baidu’s products.”

This new department was finally named as Baidu’s Infrastructure Department. Hou Zhenyu served as the chief architect of Baidu’s infrastructure department. He was among the first batch of people who really built Baidu’s technical system and the first batch of “cloud builders” on the Chinese Internet.

At this moment, Baidu paid attention to Google’s three cloud computing papers. These three papers, which are considered to be milestones in the history of cloud computing, respectively describe the distributed file GFS (Google File System), the distributed database Big Table and the programming mode Map/Reduce, which also accelerated Baidu’s move towards the cloud.

Soon, in 2010, Baidu decided to build a larger company-level infrastructure department on top of the original infrastructure department. The goal was to change from targeting non-search products to connecting all Baidu technology bases. Baidu’s company-wide infrastructure and engineering projects were officially unified for the first time this year.

2010 was also the peak of Baidu’s PC Internet era and a peak of its stock price. Baidu engineers with sufficient funds began to purchase various servers on a large scale. It is reported that the scale of servers at that time was only tens of thousands of units. With a big wave of Baidu’s hand, Baidu quickly established two computer room projects of the order of 100,000 units.

When the resources under management reached a new level and then reached a terrifying magnitude, Baidu began to consider building an underlying infrastructure that could be pulled horizontally.

One of the external aspects of the implementation of this architecture is BAE (Baidu App Engine). Google also has a corresponding concept of GAE (Google App Engine). Hou Zhenyu commented: “This may be technically different from the current PaaS. It’s the same, but the concept is the same, a layer on top of a cloud-based technology base.”

Some senior practitioners believe that Google’s GAE and Baidu’s BAE can be regarded as the first generation of cloud products with the concept of PaaS. The cloud and some underlying capabilities are used at the same level, but these concepts have not been transformed into today’s mainstream PaaS or cloud-native architecture, which is a “too advanced attempt”.

“It’s like Google and Baidu have developed microwave ovens at that time, and then used them to promote, but at this time, users don’t even have electricity at home, let alone buy ‘home appliances'”, the industry said as an example: In contrast, AWS and Alibaba Cloud are not so technologically advanced, but they follow the path of selling electricity – selling light bulbs – selling electrical appliances, which is not as cool but more in line with the underlying logic of business.”

Hou Zhenyu believes that in a broad sense, Baidu has been a cloud company from the first day; in a narrow sense, cloud computing is essentially the development of distributed computing, parallel computing and grid computing. As early as 2008, Baidu’s Hadoop distributed system had 300 machines and 2 clusters, which already had many characteristics of cloud computing.

The traces of cloudification are obvious, but for a long time, Baidu has not used the word cloud internally.

2011 was the time when Liu Qiangdong noticed that Dropbox was tempted to become a cloud, and it was also an important node when Baidu began to consider public cloud and private cloud issues. At that time, Baidu used the open cloud and personal cloud to distinguish, and the personal cloud later incubated and grew into the well-known Baidu network disk business.

At the end of this year, the mobile Internet took off. Hou Zhenyu recalled that at that time, Baidu did not use cloud purely as infrastructure, but as an integral part of the mobile ecosystem. Therefore, the decision after this important node was to set up the mobile division on New Year’s Day in 2012.

When it comes to cloud, Baidu clearly has a leading position in technology, but it did not provide it to the industry in an open form earlier, and missed the first batch of “low-hanging fruits”, that is, the first batch of Internet enterprise customers to go to the cloud. It is somewhat unfortunate that a clear strategic position has been established.

But we also pointed out in the article “Intercept” that the promotion of cloud computing requires three conditions-sufficient technical demonstration, sufficient market education, and good enough B-side sales capabilities.

Baidu is good at technology, and it is also a model of cloud use; but for B-side non-Internet companies, cloud computing is too new and too difficult to understand. Later, Alibaba, which has a stronger awareness of market operation, has single-handedly promoted the cloud. Promoted the market education of cloud computing in China.

As for sales, since Baidu was born, technical figures in the To B business have always been the absolute mainstream. In Baidu’s culture, the priority of sales capabilities will not be too high.

If Baidu attacked the cloud rashly at that time, I am afraid it would not have a good ending.

From 2015 to 2016, Hou Zhenyu believes that this is a watershed in China’s cloud computing, and 2015 is also an important time for Baidu to officially sound the horn to attack the cloud computing market.

But in these critical two years, Baidu Cloud was actually absent from the market. Although Yin Shiming joined Baidu Cloud at that time, as a typical To B sales talent, he brought many big customers to Baidu Cloud together with Zhang Zhiqi, Jing Wei and others, but the aggressive sales strategy he advocated could not last long. Still not good at government-enterprise relations, still unable to step on the market rhythm at that time.

Hou Zhenyu himself also left Baidu and went to Jinshan during this time. When he returned to Baidu in 2018, CTO Wang Haifeng had completed the reorganization of Baidu’s To B business. The banner of open cloud returned to the infrastructure department, which was later the Baidu Intelligent Cloud team. In 2020, Baidu Cloud also returned to Wang Haifeng from Yin Shiming, and Hou Zhenyu, who was in charge of technical architecture, began to be in charge of Baidu Basic Cloud.

Looking only at the ten years from 2005 to 2015, Baidu Cloud is often summarized as “getting up early and catching up late”, but after a few years, Baidu Cloud has regrouped.

At the beginning of 2020, Baidu integrated ACG (Baidu Cloud Business Group) with AIG (AI technology platform system) and TG (basic technology system) into “Baidu Artificial Intelligence System” (AI Group, abbreviated as AIG). To a certain extent, moving towards “integration of cloud and intelligence” marks the official “rebirth” of Baidu Cloud.

Stories that have not been written into official history have their ups and downs, and now they are also at the beginning of the next new era. Alibaba Cloud welcomed Cai Yinghua from Huawei, which is also a rare case of airborne M7-level executives in the history of Alibaba Group. Jingdong Cloud was brought under the command of Jingdong Technology and led by Li Yayun; the helm of Jingdong Cloud was changed from Zhou Bowen to Gao Liqiang. Baidu Cloud also took over from Wang Haifeng by Shen Shao not long ago, and the two Baidu veterans successfully completed the handover.

The dust has settled in the past years, but China’s cloud computing arena is far from the time when the dust has settled. If you want to share more exciting stories, you can contact the author of this article Zhou Lei, WeChat LorraineSummer

This article is reproduced from: https://www.leiphone.com/category/industrycloud/miqBXYbttdHRUilM.html
This site is for inclusion only, and the copyright belongs to the original author.