Amazon Web Services hosts some of the biggest Internet-based companies around including Yelp, Etsy, Netflix, Tinder, Instagram, Spotify. Amazon's cloud also hosts plenty of smaller ones: In all, Amazon's $7.88 billion web services business makes up about half of the entire $16.2 billion cloud infrastructure industry.
So what happens if the doomsday scenario occurs and Amazon's servers go down? Amazon is going to great lengths to try to make sure that doesn't happen, according to chief technology officer Werner Vogels. "These days, there should never be a case where your website goes down, whatever disaster there is--ever," Vogels told the crowd at the Northside Festival in Brooklyn last week.
Vogels said the company uses something called "highly available systems" to prevent this kind of crash. This refers to a system without a single point of failure--redundancies are built so that even if one component fails, the system can remain in operation. Vogel said the components are designed so that if they fail, they do so in a way that allows the maintenance team to observe and correct it--but an outside user won't detect the failure.
Amazon divides its servers into 12 regions around the world--from Oregon to Ireland to China. Each region's infrastructure is further split into multiple facilities in case one should go down.
The stakes are perhaps higher for Amazon than any other cloud server. AWS receives data from the Mars Rover, from the bottom of the ocean, and everything in between. The United States healthcare system even shifted much of HealthCare.gov over to Amazon's services in 2014 after high usage crashed its site on multiple occasions.
Cloud platforms as a whole have grown massively in recent years, but no one has been able to catch Amazon. AWS became the first service of its kind in 2006. The company's next three biggest competitors--Microsoft, IBM, and Google--combined have a smaller market share than Amazon's.
In November, Deutsche Bank estimated Amazon's cloud services to be worth $160 billion based on its projected revenue for the next ten years.
Companies generally use web services like Amazon's to avoid having to build their own infrastructure. Vogels noted that what used to cost a company $5 million in servers now costs $50,000 in rented cloud space, which makes it far easier to start a web-dependent company.
It's worth noting that Vogels's comments came the same week that a severe Australian storm caused Amazon's Sydney-based cloud services to go down, creating outages and lag times for companies hosted on the servers, including Domino's Pizza and Aussie streaming site Stan. Amazon blamed the hour-long outage on a bug in its software that caused its two backup power sources to fail after the local utility provider went down.
A similar incident occurred in 2012, when a storm in North Virginia knocked out Netflix, Instagram, and Pinterest for several hours.
To be fair, just a few incidents over the course of several years is a pretty good track record. But no system is infallible. While a total server apocalypse might be unlikely, Mother Nature continues to prove she can knock technology back to the stone age, even if only for an hour at a time.