This story was written by Keith Dawson for UBM DeusM’s community Web site Develop in the Cloud, sponsored by AT&T. It is archived here for informational purposes only because the Develop in the Cloud site is no more. This material is Copyright 2012 by UBM DeusM.

Bad Day For the Amazon Cloud

Amazon's cloud is experiencing pain at the intersection of northern Virginia and Kindle.

An outage at Amazon AWS and a little PR disaster involving a Kindle point up the cloud's growing pains.

As I write this, part of Amazon's cloud is experiencing problems. The AWS status page is showing yellow ("performance issue") for a few of its services in northern Virginia, where a large number of Web properties rent space and CPU cycles. Tweets involving "AWS" are crossing Twitter at about one per second at the moment, and have been accelerating. TechCrunch is reporting a cascading failure in the Virginia data center; what began as an outage in Amazon's EC2 (Elastic Cloud Compute) brought down the RDS database instances and Elastic Beanstalk services in the data center and are now causing slowdowns in ElastiCache and CloudWatch. This cascade happened in a matter of 28 minutes, according to Amazon status updates posted by TechCrunch.

One CTO tweeted: "I wonder how many mushroom clouds it would take for AWS to use their red 'Service disruption' icon rather than the 'Performance issues' one."

When Amazon's cloud services have problems, a large part of the modern Web goes unstable. Currently experiencing anything from slowdowns to blackouts are Reddit, Foursquare, Minecraft, Heroku, imgur, Pocket, HipChat, and Coursera. Also, critically, GitHub is affected -- so anywhere up to hundreds of thousands of coders could be dead in the water. (GitHub's AWS problems come on the heels of two days of DDoS attacks, of unknown origin, aimed at the social coding site.)

Previously, AWS suffered major outages in April 2011 and in June 2012, and both times people were shocked at how widely the damage rippled. Possibly these earlier outages might have encouraged some AWS customers to beef up the robustness of their code with contingencies for failover or disaster recovery. But clearly some of the larger properties on the Web are still unprepared for instability in their cloud provider.

Trouble in Kindle land
Amazon is in the news for another unfortunate turn of events as I write this. Reportedly, Amazon cancelled the account of a customer in Norway, denying her access to some 60 Kindle books she had paid for; a representative of Amazon.co.uk told her by email that this action was final and permanent, and that she would not be able to open a new account or deal with Amazon in any way ever again. There was no explanation forthcoming for why this action was taken and the woman was given no recourse. She was told only that her account was somehow "linked" with another that was suspected of violating Amazon's terms of service; the woman professes to be mystified by this suggestion.

Anyone who closely reads Amazon's terms of service for Kindle e-books might have predicted that such an incident could occur (people who worry about digital rights, such as Cory Doctorow, have been warning us for years). Of course, no one actually does read the ToS before clicking "Agree."

It is likely that this story will go much wider. It exemplifies the inherent unfairness of the current regime of Digital Rights Management, which runs so grossly against the expectations of ordinary people when they "buy" a "book" that lives in the cloud.

The long-term fallout of Amazon's bad day in the cloud could be less willingness on the part of the general public to trust anything with the word "cloud" in its title -- even if they don't fully understand what "cloud" signifies.