Programmers' comments on Singles' Day: What is honey to others may be poison to me

In 2016, Tmall’s single-day sales record was 120.7 billion. Although the data for this year has not yet been released, the peak of e-commerce traffic this year has ended. In this crazy shopping festival, people once again demonstrated the Chinese characteristic of strength in numbers and created another e-commerce myth.

Every year, the IT system architecture of e-commerce faces a huge traffic impact during the Double 11 shopping festival. The IT system of e-commerce websites includes online transactions of e-commerce orders, background management, supply chain management, logistics management, product database, customer service system, etc. All of them need to be well supported by an IT system with high reliability, high performance and good elastic expansion. This is not easy. This article will mainly describe the four key IT technologies and their importance in the system.

​​

[[209662]]

1. The “student committee” who distributes homework: CDN

How important is CDN? It is like a courier, responsible for delivering "data packages" from one place to another. The distance and speed directly affect the quality of the data stream, the smoothness of video viewing and the intensity of the pain of shopping.

With the popularity of live video streaming, CDN is now entering its golden period of development. The maturity of emerging technologies such as cloud computing, big data, and mobile Internet is providing strong development momentum for CDN. With the help of CDN, traffic requests from Shenzhen to place orders for shopping do not have to travel thousands of miles to the server in the Beijing computer room, but only need to select a CDN edge node closest to the user (Shenzhen) to respond to the user request.

China Unicom CDN Distribution

In fact, CDN is indispensable for the Double 11 Carnival every year. Analyzing CDN peaks can even directly reflect user purchasing habits. According to previous data, the "buy, buy, buy" mode started at 8 am on November 10, and at 10 pm on November 10, the first peak will be reached. After that, the traffic data continued to rise until after 0:30 on November 11, when the data began to decline, hitting the bottom at 5 am, and began to rise after 8 am until 10:30 on November 11, when the annual peak will be reached.

Judging from the current situation of the domestic CDN market, competition is becoming increasingly fierce. On the morning of November 6, the Ministry of Industry and Information Technology announced the latest five companies that have been granted CDN licenses. So far, the number of companies that have been granted CDN licenses has reached 39. After Alibaba Cloud completed the integration of Youku CDN business, the overall bandwidth reached an unprecedented 40TB, with more than 1,000 nodes worldwide, and low-price competition began, which further proves that the era of extensive economy has completely ended.

CDN is like a study committee member in the system. It knows clearly to whom the assignment should be sent, which is the shortest way, and how to ensure efficiency and quality while saving costs (TCO).

2. The “external relations chairperson” who is good at building relationships: cloud computing

There is a view in the industry that Double Eleven forced cloud computing to mature. I think this is true. In fact, the development of many IT technologies is due to the influence of strong market demand.

If we want to trace the history of e-commerce platforms on the cloud, we may have to start from the "de-IOE" (removing IBM minicomputers, Oracle databases and EMC storage devices during IT construction) proposed by Wang Jian in 2008. After all, the x86 platform has created today's public cloud. This far-reaching "de-IOE" trend has brought many benefits to Alibaba and has also had a profound impact on the development path of domestic IT. Until May 17, 2013, the last IBM minicomputer of Alibaba Group was offline in Alipay, and Alibaba's "de-IOE" strategy achieved phased results.

​​

[[209663]]

(1) Most data centers now use X86 architecture servers

However, not every company has as strong financial support as Alibaba's "de-IOE" strategy. For small and medium-sized enterprises, the best way to "de-IOE" is to move their business to the cloud and use the public cloud to change the enterprise IT architecture.

For a high-concurrency node like Double Eleven, the elasticity and scalability advantages of cloud computing can be fully utilized. Public clouds can fully utilize resources and reduce the company's TCO by establishing a computing resource pool and flexible software regulation.

​​

[[209664]]

(2) Cloud computing enhances IT flexibility

In order to ensure that the server does not crash during the Double 11 shopping festival and allow consumers to enjoy a better shopping experience, the IT department of the e-commerce platform will generally conduct stress testing on the existing network before the Double 11 shopping festival and make a large number of system reservations. Judging from the user experience in recent years, the IT support for Double 11 shopping festival has indeed been getting better and better.

The power of one person is always limited, and the same is true for IT systems. Cloud computing is like an external liaison chairman in the IT system, which can always help solve the problem of insufficient capacity of small groups and improve the elasticity and scalability of IT systems.

3. The “squad leader” who distinguishes between the primary and the secondary: distributed database

According to the data released in 2015, Alibaba's self-developed OceanBase distributed database carried 100% of the transaction traffic on Singles' Day, creating 140,000 orders and completing 85,900 payments per second (47,300 in 2014). Taking the transaction system alone as an example, the amount of data written into the database on Singles' Day reached 10TB.

As mentioned earlier, the original inherent database is generally Oracle's Oracle database. After "de-IOE", Alibaba developed its own OceanBase database, which is also China's first database with intellectual property rights and the first distributed data relational database used in financial business.

​​

[[209665]]

(1) Distributed database platform (currently used by Ant Financial)

The reason for building a distributed database is that Alibaba's transaction system requires strong scalability and availability. To this end, Alibaba has built a distributed C2C business suitable for e-commerce platforms such as Taobao, adhering to the principles of system scalability above all else, system availability above consistency, and appropriately relaxing consistency constraints. The combination of distributed database and hybrid cloud (public cloud + private cloud) architecture has made Alibaba's IT architecture a benchmark.

For online shoppers, there is a problem that you may be accustomed to: consistency issues. For example, you have successfully placed an order and paid for it, but the backend shows that the shipment failed. This is a problem with database consistency. In order to put database scalability first, distributed databases naturally encounter some consistency issues.

(2) “AP” is the choice of most distributed databases

In 2000, American scientist Eric Brewer once said: When the technical architecture evolves from a centralized architecture to a distributed architecture, it will encounter the bottleneck of the "CAP Law". CAP is the abbreviation of consistency, availability and partition tolerance. Eric Brewer believes that a data processing system cannot meet the three elements of consistency, availability and partition tolerance at the same time, and can only meet two elements at most. For distributed databases, it is "consistency" that is sacrificed, so we have encountered the situation where payment is made but delivery is not seen.

You can only gain something if you give up something. For a class monitor, he often has to choose between studying and class affairs, and between collective honor and personal interests. A distributed database is like a class monitor who has already distinguished between the primary and the secondary and made a choice, putting scalability first, even if it means sacrificing consistency.

4. The “teacher” who checks homework: APM

About two years ago, enterprises began to realize the power of mobile terminals, countless software companies were established, and application software became an important channel for enterprises to acquire customers and disseminate information. The importance of customer application experience was once again put on the agenda, and APM flourished in this context.

​​

APM architecture diagram (picture from the Internet)

APM is the abbreviation of Application Performance Management. It only monitors and optimizes the key business applications of the enterprise, improves the reliability and quality of enterprise applications, ensures that users can get good services, reduces TCO (total cost of use), and brings more benefits to the enterprise. At the key IT node of Double Eleven, APM can help enterprises better understand their IT advantages and limitations.

According to the APM report of market research organization Gartner, although there are large companies such as IBM, Dell, and Microsoft, there are also many SaaS companies dominated by APM in the leaders quadrant, and most of them are foreign companies (Note: Chinese company Tingyun was listed in the 2015 Magic Quadrant).

​​

APM Application Performance Management (Picture from Tingyun)

Through the previous introduction, everyone has understood the complexity of a powerful IT system. It must include multiple layers such as servers, storage, networks, operating systems, databases, middleware, web services, pages or terminals, etc., which is like a series circuit. Any link in this long chain may fail, and any performance bottleneck will affect the performance of the entire application.

The role of the APM performance management system is to provide an active management solution. In the face of a loaded IT environment, it locates the root cause of the problem through real-time performance detection and analysis, and optimizes performance. User experience is particularly important during the Double 11 traffic peak, and no mistakes can be made. Taking Tingyun as an example, based on previous Double 11 experience, it has developed a strategy of preventing accidents, quickly locating and quickly solving problems, that is, a trinity of prevention, diagnosis and treatment.

However, it is worth noting that in the past two years, with the maturity of microservice architecture and container technology, IT systems have become more flexible, but IT environments have also become more complex and more difficult to monitor. This requires APM performance management systems to provide sufficiently fine-grained monitoring suites. In the new environment, how to track the key to performance problems in information links and system architecture topology has become an urgent problem that APM vendors need to solve in the future.

Double 11 is a big IT exam. In this exam, APM is like a teacher, helping candidates prepare in advance, monitoring the exam situation, and making a final judgment. Of course, the purpose of this teacher's invigilation is not to catch cheaters, but to help candidates pass the exam.

5. “One man’s meat is another man’s poison”

Since the concept of "Double Eleven Shopping" emerged, e-commerce platforms have been better and better every year, sales have been rising year after year, and people seem to have gotten used to breaking records every year. At the same time, IT technology has been constantly improving, with increasing scalability, availability and reliability. There are fewer and fewer incidents of page crashes caused by server downtime during high traffic periods. In response, e-commerce platforms are purchasing more and more IT equipment (servers, storage, etc.), and the pressure on programmers, IT engineers and architects is increasing.

​​

(1) What is honey to one may be poison to another

Jack Ma once said: "Through Double Eleven, we can truly improve the technology and level of China's future new retail, because the improvement of this technological capability is really rare. At the same second, so many consumers come in. This technology, the breakthrough in cloud computing capabilities, and the breakthrough in management capabilities, I think are amazing." Every year's Double Eleven has obviously trained the IT department's ability to handle high concurrency and helped the continuous improvement of the CDN industry.

However, opportunities often coexist with challenges. The emergence of Double Eleven has indeed promoted the rapid development of online e-commerce. In order to support the huge technical needs of e-commerce, IT engineers, programmers, and architects have been expanding and upgrading IT systems day and night to improve IT capabilities.

On the other hand, in the context of cloud computing, big data and artificial intelligence, the large amount of data generated by the Double Eleven e-commerce festival cannot but be called corporate wealth. By storing, organizing and analyzing this data, more valuable marketing information or sales leads can be obtained, which are also necessary conditions for implementing precision marketing.

​​

[[209667]]

(2) As a shopaholic, I would like to thank the IT staff

"One man's honey is another man's poison," said the programmer. At first, I didn't understand the meaning of the programmer's sudden words until I learned the above technologies and found that the Double Eleven Shopping Festival, a national carnival shopping festival, is just an absurd drama composed of countless codes and 0s and 1s in the eyes of a technology geek. The leading actors of this drama are hundreds of millions of netizens, the director is the e-commerce platform, and the field staff, ticketing staff, sound engineers and lighting technicians who really provide support are these programmers. Thank you for your hard work, programmer~