Well...
The news about the Xi'an health code system collapsing twice in a row has been fermenting on the Internet for several days, and Presumably the poor friends almost all know it.
The head of the department responsible for the construction of the health code platform was also suspended yesterday, which is an explanation to everyone.
However, just today's bad reviewer found that Xi'an's health code was rushed to the Zhihu hot search.
The reason is actually not complicated, that is, in June last year, a report on Xi'an's "one code pass" (health code) was turned over by everyone.
The content of the report is almost that the members of the one-code team have endured hardships and hard work, worked tirelessly, solved many dangers and obstacles, and played a huge role in the prevention and control of the epidemic.
The report is not too long, and interested partners have time to search for themselves.
If the health code in Xi'an had not collapsed twice a few days ago, then this report can be written no matter how it is written, and no one will care.
" With a yard pass, epidemic prevention is easier " ▼
But at the stall of epidemic prevention and control, Xi'an's health code gg has been.
So this article has become a technical data for everyone to analyze "why the one-code pass collapsed".
As a result, after reading it, you found: Oh ~ It turned out that you spent two days and two nights compressing the 1 MB picture to 100 KB.
The size is only 1/10 of the original, which is so powerful.
Poor friends who know a little bit about technology may have heard that I am yin and yang weird.
Yes, I'm just weird in yin and yang.
It's not that I'm blowing, don't look at me, I'm a stinky writer, but it only took me 30 seconds to think of two ways to compress the size of a health code picture tenfold:
1) Use jpg/webp/heif and other formats to recode health codes;
2) Draw health codes using svg vector graphics;
The first method is actually what everyone will do, that is, to find an image compression tool to transcode the picture.
Don't say 100KB, 10KB can be pressed to show you.
The second method is to write the image elements in HTML code.
The final file size is also smaller than that of a normal image.
And most critically, both of these methods have mature code to copy.
Do a casual GayHub search, whether you want to run on the server or in the browser, just copy and paste a few lines of code to paste.
One minute, what supporting library supporting dependencies are arranged for you, how can it be two days and two nights in front of the computer to do this...
Hundreds of jobs to choose from... ▼
And if the team of "One Code Pass" is really so independent and pure hand rubbing, I know that the big guy DBinary also helped them try it.
The result is a person, a pair of keyboards, an hour and a half...
So if Xi'an's one-code team really spent two days and two nights studying image compression...
I couldn't understand it, but I was blown away.
And in fact, enthusiastic netizens with good things have also done element audits of the web page activities of Xi'an Health Code.
It turns out that the health code itself is transmitted in the form of a string, and then the QR code picture generated locally from the string is generated on the phone.
The string itself transmitted from the one-code pass server will not exceed 1KB at all...
As for the image compression mentioned in the report, the high probability of compression is the two in the screenshot below...
And not the health code itself that our brain supplements.
Actually... As everyone said, the authors who wrote this report are most likely not exposed to technology.
For example, I don't know if the bad friends remember the video encoder that We talked about last year on Alibaba's "world's first"...
So it's normal for the reporter to send out an article that looks powerful, but actually can't poke the technical pain points.
It is likely that there may be a certain technical cost when the one-code team locates this picture load problem, but when it is implemented to the pen, the reporter only gets to the first layer...
But these are digressions.
So why, in the end, Xi'an's health code will collapse twice in a row?
I also chatted with a few friends who had similar project development experience, and finally everyone came to the conclusion that the database did not resist.
Because according to public reports, Xi'an's one-code platform has expanded the network export capacity to 10 Gigabit last year, and it may be higher this year.
There were only eight servers at the beginning, but in the past year or so, One Code has also invested millions to purchase cloud service facilities.
It can be said that from this scale point of view, even if you really have 1MB in a picture, even if the static resources and interfaces are placed on the same domain name, hundreds of thousands of people access up to multiple card points at the same time, and the entire system will not collapse.
But the database thing is very metaphysical...
Its carrying capacity per second is related to the size of the form, the standard, the read and write speed of internal and external memory, the computing speed of the host processor, and the throughput capacity of the network.
Otherwise, Oracle (Oracle) Company in the United States can not rely solely on the development of database software and solutions, and has consolidated itself into a global top 500 with a market value of trillions...
In particular, the digital infrastructure of the health code accommodates almost the population information of the whole city, plus it itself has to be linked to the health code system at a higher level, and the database system of vaccine data, nucleic acid data, trip data and so on.
The logic behind pulling a health code is not uncomplicated.
If the database of the health code is not selected at the beginning, and the isolation is not done well, then once the load is loaded in the later stage and the load bottleneck is found, the probability can only be solved by burning money to expand the performance of the heap server.
( However, the above is only a bad reviewer and some small partners speculation, if there is a small partner has more accurate inside information, welcome to punch the face ~ ) )
In fact, the bad reviewer has talked so much with everyone, and there is a little bit of afterthought.
After all, at that time, all the teams that developed health codes in China were the first time, and the performance of the products made without the lessons of the past was lower, which was also understandable.
However, in terms of the health code collapse of the two full nucleic acids, it is obvious that Xi'an One Code Pass does not have an accurate estimate of how many people will call the service at the same time, and the expansion is not in place.
Why this otherwise avoidable futility could have happened is something I do not understand.
In the past, when there was an unpredictable event such as a celebrity divorce on Weibo, the server had to be shut down by the melon eating masses.
After that, Hu Zhongxiang would run out and say, Ah too sudden, we are already in an emergency expansion.
Within half an hour, Weibo's service will begin to gradually recover, and everyone will continue to do what they want to do.
But the health code is not a microblog, and the nucleic acid of all employees is not an unpredictable event.
This pot, Xi'an's one yard pass or a good back.
Hey, say a thousand ways and ten thousand, now everyone hopes that in fact, xi'an's one-code team can learn the lessons of these two times.
After all, the health code of "helping to fight the epidemic" should not be so jumpy and become a resistance on the road to the fight against the epidemic.
Author: Miro Editor: Xiao Xinxin & Jiejie
Resources:
I know that Xi'an Telecom One Code Pass Project mentioned in a previous report that "two days and two nights optimized 1m pictures to 100kb", what is the difficulty of image compression technology? , the answer of the cat-owning husky, DBinary and others