On the day of the Chinese college entrance examination, most of the Internet in the world was paralyzed by this "small company"

On the day of the Chinese college entrance examination, most of the Internet in the world was paralyzed by this "small company"

On June 8, while the college entrance examinations were in full swing in various provinces and cities across China, something major happened on the Internet abroad.

On that day, hundreds of millions of Internet users found that they were unable to open the websites they normally visited.

[[405149]]

  • The UK government's main public service portal is inaccessible
  • A large number of news websites, second-hand trading platforms, and online shopping platforms, including the British "Guardian", the American "New York Times", the French "Le Monde", the Japanese "Yomiuri Shimbun", the Financial Services Agency, and the online TV station AbemaTV, all experienced disconnection.
  • Many websites including Google, Amazon, PayPal, Twitter, etc. experienced "503 Errors" and the websites could not be opened.

The outage affected three continents: Europe, Asia, and North America. Amazon alone lost approximately $34 million in sales.

[[405150]]

It took from a few minutes to an hour for all websites to return to normal. After investigation, it was found that the culprit that caused most of the global Internet to be paralyzed was a company called Fastly.

Who is Fastly?

Fastly is a relatively unfamiliar name to ordinary Internet users. It is an American cloud computing service provider whose edge computing platform mainly provides content distribution network, network security services, load balancing and video streaming services. It is headquartered in San Francisco, California, USA, with branch offices in Denver, New York, Portland and Tokyo, Japan, with more than 900 employees.

The company's name sounded unfamiliar, and it had only more than 900 employees, which made many people think it was a small company.

In fact, Fastly provides cloud services to many well-known companies, such as CDN. Fastly builds a relay station between the server and the end user. It will first cache the resources on the server to its own node. When the user initiates an access request, the node closest to the user will be responsible for responding and sending the relevant resources to the user.

CDN can significantly improve the loading speed of web pages, optimize images, videos and other Internet resources, and present them to end users faster.

The inaccessibility of several well-known websites was caused by a bug in a software update sent to customers on May 12.

It was not until June 8 that the customer changed the service configuration, triggering the bug, which eventually led to the downtime of a large number of well-known websites.

Fastly said that the problem occurred at 9:47 a.m. Greenwich time on June 8 (17:47 Beijing time). The company's engineers noticed the problem within a minute and analyzed the cause in about 40 minutes. After disabling the settings that caused the problem, most websites were able to return to normal.

After the accident, people discovered that this little-known "small company" actually provided cloud services to so many well-known companies. On the same day, Fastly's stock price rose instead of falling, surging 10.85%.

After seeing this news, many people couldn't help but wonder: It is understandable that PayPal, Twitter, and major news websites use Fastly's CDN service. After all, they don't have it and can only use others'.

But Amazon has AWS and Google has Google Cloud, so why would they use a third-party CDN?

[[405151]]

Generally speaking, large companies will choose CDN products from multiple cloud service providers. The main reasons are as follows:

1. Different coverage areas

Each CDN service provider covers different geographical locations, and using a single service provider's CDN cannot meet the high requirements of large enterprises.

2. Different product features

In addition to providing acceleration services, the functions provided by different CDN manufacturers may also vary. Some support clearing built-in SSL, routing rules, etc., while others support VoD or provide real-time streaming media, etc.

The Fastly outage incident sounded a wake-up call to many cloud service providers. If a short-term failure occurs, large-scale Internet services will be affected, and such incidents should be avoided as much as possible.


<<:  Share an interesting data analysis method

>>:  Read this article only three times and you will never forget network layering!

Recommend

Ruijie appeared at GITC: Internet companies please give me a thumbs up!

[[177701]] [51CTO.com original article] The GITC ...

Will 5G really kill Wi-Fi? NR-U is here

5G is like an elephant running wildly in the wire...

If you don’t know NAT, you can’t be a good network engineer

[[436174]] 1. Network Address Translation Technol...

Another batch of long transactions, who is to blame for the P0 failure?

In recent weeks, there have been many service err...

Problems that edge computing needs to solve urgently

At present, edge computing has been widely recogn...

Monitor infrastructure to prevent unexpected downtime

[[258649]] 【51CTO.com Quick Translation】Infrastru...

6 solutions to the problem of no Internet access in Docker containers

Note: The following method is a solution that all...

As of December 2020, there are 229 million 5G users worldwide

On December 22, according to foreign media report...