Is it technically possible to pull all 1.4 billion Chinese people into a WeChat group?

Is it technically possible to pull all 1.4 billion Chinese people into a WeChat group?

Recently, Zhihu has a very hot question: "Can it be technically possible to pull 1.4 billion Chinese people into a WeChat group?"


In fact, according to the National Bureau of Statistics, as of the end of 2017, the total population of mainland China was 1.39 billion (including 31 provinces, autonomous regions, municipalities directly under the Central Government and active military personnel of the Chinese People s Liberation Army, excluding Hong Kong, Macau, Taiwan and overseas Chinese). Number), has already exceeded 1.3 billion.


At present, the upper limit of the number of WeChat group members is 500, which draws nearly 1.4 billion Chinese people into a WeChat group. Do you consider reality from a technical point of view? How many servers are needed?


And in a group of 1.4 billion people, how to grab red envelopes? Will the phone explode? Welcome everyone to watch the official Tucao series~


Let me start with the conclusion: it may be possible, but you will see nothing.


According to the public data of the 2017 WeChat Data Report: In September 2017, WeChat had an average daily landing of 902 million people, and an average of 38 billion messages were sent every day.


This means that the average person sends 42 messages a day. If the people of the whole country (by the way, the population of the country is now close to 1.4 billion) speak in the same group, the number of messages that appear in this group every day is as high as:


If so much information is only sent at a constant speed, taking into account everyone s sleep, the 8 hours of sleep is not counted, then the information to be received in the mobile phone every second is:


Wow, over 1 million per second! At present, one of the mobile phone CPUs with the highest clock speed, Qualcomm Snapdragon 845 has a processing capacity of 2.8GHz, with a total of 8 cores.


If CPU operations such as Android system, display refresh, and network IO are not calculated, the computing power that can be allocated to each piece of information is:


What is this concept? The world's first microprocessor was Intel 4004, which was launched by Intel in 1971. The clock speed of this old antique is also 108KHz. So 21.9KHz can't do anything.


Fortunately, there is a Moore's Law in the IT world: CPU performance can double every 18 months (or half the price). Although the current technology has been difficult to increase the frequency (a toothpaste factory desperately squeezed only 5 Ghz).


But suppose we use black technology to increase the clock speed. When Moore's Law fails in 2025, the CPU frequency of our mobile phone should reach:


It looks good, but the computing power available for each message will reach:


Haha, it still hasn't reached the level of Intel 4004, so the result is that you have waited 7 years and still can't get into this national group to grab a red envelope.

Well, let's connect the mobile phone to a powerful computer, such as the world's number one supercomputer Taihu Light, with 10 million CPU cores to help deal with the largest WeChat group in the universe. The problem of computing power has finally settled.


We assume that each message has 10 Chinese characters on average, which is roughly equivalent to 30 bytes. In addition, the application layer will add certain control characters, and the data consumption of the TCP/IP network layer is about 74 bytes, rounded up. On average, each message has 100 bytes, and each byte is equivalent to 8 bits.


At this time, the network bandwidth required per second is approximately:


If someone sends a red envelope, the bandwidth needed is even greater.


In theory, 4G networks can support 1000 Mbps, but don t forget that the people of the whole country are in the same group, and people around you also need the same bandwidth, which makes the base stations near you overwhelmed and paralyzed.


In order to avoid network paralysis, you need to move to a base station where no one is around, such as when you are the only one in the school during the summer vacation.

However, the life of the operators is not easy, because the traffic across the country has reached astonishing in this second:


This is equivalent to 65.7% of the total national mobile data traffic in April 2017, which means that every 18 seconds can use the entire country's traffic for a year.


If 1.146 Ebit data is installed with a 2TByte 3.5-inch hard drive (20 mm high), and then stacked, it will be 1433.25 m. In contrast, the world's tallest building, the Burj Khalifa in Dubai, is only a mere 828 m.


Of course, if there is a need, I believe that telecom operators will definitely spend a lot of money to build the world's largest broadband network for you.


However, it is not the operators that should spend money next-but Tencent.


In order to handle this 1.146 Ebps traffic, Tencent needs to prepare 114.66 million sets of switches and servers.


At present, the price of a large factory 4-port 10 Gigabit switch is about 4,000 yuan, and a cheap server with a 10 Gigabit port costs about 10,000 yuan. The combined cost of these two items is:


Well, these two items alone are equivalent to Shenzhen's GDP for the whole year of 2014.


This does not include network cables, wires, server racks, computer room hosting, electricity bills, operating expenses...


The storage of so many devices is also a problem. A 2U server with a 10 Gigabit (10Gbps) port is 88.9 mm high, so when stacked, there are:


This is almost the distance between China and the United States, and it is enough to repair railways.

Well, with so many device blessings, you can finally join the group happily.


But you are surprised to find that there is nothing on the screen except white-this is because your eyes cannot receive such fast data!


The visual persistence time of the human eye is 100-400 milliseconds, and our group has to display 1.02 million messages per second, and the dwell time of each message is only about 0.001 milliseconds. In contrast, movies and TV have 41 milliseconds.


So before you have time to read the message, it has disappeared, leaving only a white patch in the center of the screen.

Here are some comments from netizens:

@ :

1.4 billion in a group is not terrible. The scary thing is that every holiday group will let the group owner give out red envelopes!


@ :

It has been done. 1.4 billion people have been drawn to a WeChat group, and all they have seen are news broadcasts.


@bluecat:

Simply put, your phone will crash immediately because it cannot carry the amount of information for a second.


@

It can be achieved, but there are a few restrictions:

All WeChat accounts are mandatory to join this WeChat group.

The WeChat group can only speak for a limited number of people, and other people cannot speak.

The WeChat group can only send messages at a fixed time every day.

Other WeChat groups cannot send messages at a fixed time, or can only forward messages from this WeChat group.

This can be achieved without technical difficulties.


@ Morgan

"Pulling" to a group is not difficult to achieve. Anyway, the user information is on the server, and building a group that includes all users' WeChat accounts is just adding a record.


However, this group must not allow anyone to speak, just because of the diversity of our people, all kinds of chats, map madmen, advertising madmen... Massive information can instantly connect servers, operator networks and The battery of your phone has crashed.


@

To tell the truth, the current technology is still feasible theoretically, ahem, I'm going to pretend to be forced.



After reading the answers of other respondents, they said that people, terminals, transmission, processing, storage, analysis, etc. have defects or shortcomings, and cannot keep up with the large quantities of data. In fact, I feel that it is still necessary to implement it. Yes, it's just the relationship between cost and profit.


First of all, you have to consider the human factor. How many billions of information are worth infinitely close to zero for a specific individual. I personally don't pay attention to this information at all, because the efficiency of obtaining information is too low.


This has caused 99% of people to directly ignore the existence of this group, and the rest of the daily news in this group is nothing more than top announcements, top news, red envelopes and idle chats, browsing announcements and news.


Considering the problem of concurrency, the current server can do it, after all, there are a lot of news apps that can do it; red envelopes, make an algorithm for random allocation, don't grab it, grabbing will seriously affect the experience, and give one billion It should not be too difficult for users to randomly allocate a piece of data.


The rest is chattering with pictures, the data is stored directly on the server side, and the analysis and processing summarizes the central idea and pushes it to individual users every few seconds. It is almost the same. If you need detailed information to search on the server, I think it s personal. The pressure on the terminal will not be too great.


Secondly, transmission. This is the link that I think is the least problematic. Why? After solving the problem of personal terminals, the amount of personal data transmission is not large, and the existing transmission network can fully meet the requirements.


Server-side transmission depends on how the server is built. If centralized processing and storage are used, only 100G dedicated lines can be used, and three to five lines are completely enough.


It's just that the corresponding supporting switch router needs to build a huge system. If it is distributed storage and processing, 10G or even GE dedicated lines are sufficient. This is transmission.


3. processing. If you have to centrally process large quantities of data, you have to build a set of the largest and even the largest and most complex data center in the world to be able to host this system.


But if distributed processing is used, I believe that the current system is also sufficient. After all, the amount of existing calculations is already so large, and the amount of data will definitely not increase exponentially after this group.


4. storage and processing work can be completed. Storage is certainly not a problem. It is even possible to store data on personal terminals after fragmentation, and transfer the contradiction of investment equipment to data security and management.


Fifth, data analysis. This is the most important point. Among the difficulties, how to effectively analyze and extract useful information from such a large amount of data and push it to specific individuals is the core key.


Although the technology has not yet been commercialized on a large scale, I believe that this technology must have been tried or even commercialized, but the public is not clear about it. After all, this kind of thing is still a bit scary to think carefully.


In short, how to implement this system or build this group is nothing more than to transfer the contradiction between demand and resources, to transfer the contradiction between large storage demand and large investment to data security and operation management, and to transfer big data. Mass transmission is decentralized, a large amount of data is analyzed and extracted, and then directed to push, the core investment is a complete set of intelligent and effective big data analysis system.