Why would Cheng Xuyuan say that banana success is a miracle?

In the meantime, Lu Zhou naturally understood the truth in Cheng Xuyuan's words. This originated from the methodology of product iteration and update.

After all, Qiandu is one of the three giants, and Cheng Xuyuan worked at Qiandu for a while.

If anyone who knows what Lu Zhou knows about the advertisements for Menggu and Fang Qiang, he will definitely shock the chins of a group of people.

Luzhou's product sense of smell is just a little bit, and it's right as soon as it is done, and it's almost a monster. This is not something that ordinary people can do.

The most important thing is that Banana was successfully made, and it raised a lot of money, and it is still expanding on a large scale.

Having said that, Lu Zhou also knows how to update the subsequent version of this product once an Internet product is released? There are a lot of different means and strategies in this field.

As for Menggu Advertising, all the initial architecture and product prototypes came from Lu Zhou’s own ideas.

At the beginning of the business, Menggu Advertising was built with a one-year lease for ten servers provided by Phaeton. In the early days, these servers were not even fully loaded.

The business volume in the early stage was relatively small, and Luzhou naturally did not follow the conventional development path.

Now, there are about twenty servers carrying the daily operations of Menggu advertising.

This is the business expansion brought about by Banana since its launch. Most of the businesses in the middle have also been gradually divided into microservices in subsequent version updates.

Most of these microservices are balanced through a load balancing service to access users. When they are nearby, they will be idle for users to access data on the corresponding server.

If Fang Qiang wants to update the Dream Valley platform or Banana Market now, how should he do it?

Assuming that the 2.0 version is running on the server today and the 3.0 version will be released tomorrow, then Fang Qiang has to log in to the server in the early morning and update the version on the server to 3.0 through remote instructions or scripts.

In short, Menggu Advertising is now the stage of "violent release".

When Fang Qiang decides to upgrade the version of Menggu Advertising, the code versions of all services on the twenty servers will be replaced from 2.0 to 3.0 in a certain centralized time. During this upgrade process, related services will also face interruptions.

Even once a sudden bug occurs in the new version, the server will have to face the problem of version fallback.

Naturally, "violence" has the benefits of violence. It simply and roughly relies on manpower and scripts to solve daily iterations, which are simple and cheap. This is very suitable for startups like Menggu Advertising.

Compared with "violence", the more elegant one is "canary" or "grayscale release".

The word canary comes from the practice of mine workers in Britain in the 17th century carrying canary down the well. Because it is difficult for human noses to smell the smell of gas. However, canary is extremely sensitive to gas and can warn whether there is gas leak in the mine.

The release test on the Internet refers to "canary release", which means to use some of the servers or services as "canary" and "villain" as the original "canary".

If Dream Valley Advertising is to replace this version release mode, it is to update the 3.0 version to one of the "Canary" servers during each update period.

After Canary confirms that the new version function is normal, the subsequent server version will be released one by one.

Of course, when thinking about this, Lu Zhou felt that this was not as mysterious as what he said. It was not so much an iterative methodology, but rather a normal means for ordinary people to deal with things.

Throwing a stone to ask for directions is nothing more than this.

Lu Zhou smiled and said, "Don't let me lie to the big engineers who come out of Qiandu, this temple is a small one. In order to invite engineer Cheng to come, I'm going to sell the blame."

"Selling the bad guy? Are you teasing me? I'm going to calculate how much annual salary Qiandu will give you, how much annual salary you give you. If it weren't for the same bed in the past, I wouldn't have been happy to come."

Lu Zhou didn't take on this joke, "I'm old, don't mention such sour things. In a nutshell, everything is going to be done. I don't have the fancy things you use in thousands of degrees. I have to start over with honestly and step by step."

Cheng Xuyuan was not surprised, "Okay. I kept promoting the development progress before, but after all, I didn't go online, so I really didn't pay much attention to this. I thought Vancouver was doing this for you. If you don't have this, just write it again. It's really not a problem, it's just a waste of time."

Lu Zhou continued, "Well, it's not too much trouble. It's just that time is a bit tight and there are not enough people. In the past or two weeks, I'll be rubbing a wheel. By the way, the Qiandu A/B test should be used a lot, right?"

"Well. Functional switches, a/b, grayscale, etc. are used in daily life. Some old projects may use shadow testing. Very process-based things, depending on different project groups, they will generally be used to build a separate management platform."

Lu Zhou touched his chin, and after a moment, "Give you some imagination space. I won't mention anything else. How about a/b testing being used as a cloud service? It's not that I'm crazy about money, but just doing this thing, there's no need to repeatedly create wheels inside."

"Hmm?" Cheng Xuyuan was shocked and was surprised.

It is not accidental that Luzhou has this idea.

Whether it is a/b testing or grayscale release, it is correct to adhere to the same principle as the market and users.

Lu Zhou has a prophet in his mind and some prototypes about WeChat in his mind, but is it absolutely correct? Lu Zhou is willing to believe that his memory is not wrong, but he is also unwilling to bet on its correctness.

His ideas can set the direction, but they cannot represent the real needs of other users.

Specifically, when it comes to how Luzhou should formulate a WeChat release strategy and write a software platform, it is still a "sending stone to ask for directions".

At present, WeChat has initially realized the first stage of mobile communication. The version of this stage carries the most basic functions of WeChat, such as address book, sending and receiving messages and pictures, voice and video functions, lbs, etc.

At this stage alone, Cheng Xuyuan has successively made several versions of product depth.

Under Grayscale Release, WeChat can completely publish applications to specific users through the control of the publishing platform for users to use. After obtaining feedback from these users, Luzhou can combine operation statistics, etc. to determine whether the functions of a certain version are suitable for large-scale release.

In this case, Luzhou can choose the channel it has already connected as arable land. According to the user portrait, push the WeChat message to banana users of different groups as the initial user of grayscale release.

Similarly, gray tests are similar. In subsequent small versions, first test to a small range, and then fully release them after obtaining correct feedback.

Just like the Qiandu Internal mentioned by Cheng Xuyuan, this is a process-based thing. Large Internet companies will formulate management platforms as supplementary feedback in addition to data statistics.

It first processes the process of product improvement and iteration, and then uses technical means to realize the so-called methodology.

Everything is not black or white, but just gray. From the release to the test stage, every new thing released is just a canary and a stone asking for directions.

There is no absolutely correct product, only absolutely correct data and users.
Chapter completed!