Amazon Web Services had a problem yesterday that caused a high error rate on its East coast customers.
For people who don’t follow the behind the scenes stuff on the Internet, Amazon is the largest provide of “cloud computing” services. Back when they were just selling books, they discovered that web hosting for their own business was too expensive, complicated and broke a lot, and was not easy to expand as their business grew, so they invested money in finding solutions. They came to realize they were not the only company with those problems and decided to offer a service to others, which they started offering in 2006
The disruption on February 28, 2017 was in their S3 service – Simple Storage Service.. The dinosaur media is full of “the sky is falling” news stories that seem to reflect an aspiration that they can convince the world that the Internet should go away.
- “the Internet went down/broke” – No, it did not. That is about like saying because a truck crashed on the George Washington bridge in New York, that there were massive traffic jams in Paris. The roads of the Internet were not affected and most people did not notice the problem at all
- “Amazon Web Services is down!” – not accurate. Only the customers using the S3 service out of one data center were affected. The service never went down, but it did get slow for a few hours as the underlying problem was being resolved. Amazon has datacenters all over the world. More importantly, no data was lost.
- “There goes their 6 nines reliability rate” – not true. That’s not how outages are calculated. Even if you accept the premise that this was an outage and not a degraded performance, the outage has to be expressed as the portion of the S3 service that was affected. While Amazon has not been specific, probably around 97.62% of Amazon’s S3 customers were not affected. Amazon’s service is provided out of 14 regions around the world.
While Amazon hasn’t said this, the list of affected sites suggests to me they were customers who saved money by using S3 Reduced Redundancy Storage. The S3 service will create multiple copies of data and immediately access it from another region, but customers of this service accept the risk of an outage and are only guaranteed 99.99% uptime, or being down about an hour a year.
From the archives of South Park It’s just gone!