INTERNET KA DABBA GUL – THE FASTLY FIASCO
Released on : 2021-06-12
INTERNET KA DABBA GUL – THE FASTLY FIASCO

The Outage

On Tuesday (8th of June) a number of most visited websites went offline and returned an error number 503 when users tried to reach those sites. This error means that the browser was not able to access the server. The major sites which got affected by the outage, as reported on Downdetector. in, were The New York Times, Financial Times, The Guardian, CNN, Reddit, Amazon.com, Spotify, Twitch, Pinterest, Stack Overflow, Github, gov. UK, Hulu, HBO Max, Quora, Paypal, Vimeo, and Spotify. The users could not access the sites for about 50 minutes or so. At first, it was speculated that the outage was the result of a serious cyberattack but later it was traced out to be an issue with the content delivery network (CDN) of Fastly.

What is Fastly?

Fastly is an American Computing Service Provider and one of the largest CDNs. The company was founded by Artur Bergman in March 2011. It got listed in 2019 and is having a valuation at around $6bn. Other big names in this field are Amazon Web Service (AWS), Cloudflare and Akamai. In terms of value, Fastly is much smaller than AWS which is worth 13.4bn and accounts for a third of $42bn cloud market.

Role Played by Fastly?

The CDN of Fastly acts as proxy servers which are physically closer to the end-user. The web pages are cached at these proxy servers for faster and optimized delivery. In simpler terms, say if the users are to load a web page every time from the server where it was initially uploaded, it takes around 100 milliseconds to access the page. But if that web page is cached at a proxy server, closer to the user, then it would take just 1/4th of that time because the length of physical barriers is reduced.

This way companies like Fastly cut loading time for images, help faster delivery of websites and apps, handle traffic spikes and Distributed Denial of Services (DDoS) attacks etc. on the websites, help to accelerate Application Programming Interface (APIs) to improve performance and allows to make real-time configuration changes for websites and applications.

Core Problem Behind the Outage

The outage problem was triggered in the afternoon. All the above-mentioned websites reportedly went down. Within one minute it was found that the outage was due to a flaw in the configuration of Fastly’s CDN. In the next 48minutes, 95% of the company’s affected network was brought to normal, according to Fastly.

It has been reported that Fastly began a software deployment in May. But there was a bug in its code which stayed dormant till 7th June. On 8th June a single user pushed a valid configuration that included specific circumstances that triggered the bug. The configuration took down 85% of the company’s network, according to Fastly.

It needs to be mentioned here that this is not the first time that such a problem came into the picture. In 2020, two major outages were reported on the Downdetector website. One was in July 2020 when a major disruption in Cloudflare’s service broke the internet. Multiple services went down on the 17th of that month. The second one happened on 25th Nov 2020 when Amazon’s Cloud Service got disrupted. Most users complained of problems with the AWS console. The outage affected users in India, Japan, the Netherlands, and the USA.

Technology is Alternately a Boon and Bane

Suppose on a very normal working day you are doing your regular stuff on a website, reading news, or trading and suddenly everything disappears from your browser printing a weird-looking error. It just reflects the fragile nature of modern technology. Just a miniature problem or flaw at one point of the communication channel between the servers and the end-users can become utterly obnoxious and costly. It could have a huge impact on the business especially e-commerce websites like Amazon. Some media reports have mentioned that for each second of an outage, Amazon.com would potentially lose $6803. In the age of the internet even for small negligence, repercussions can be huge. However, quick detection and isolation of the problem within a vast network also reveals the robustness of modern technology.

Author: JJSingh