Amazon Down Briefly Sunday in Europe

Amazon's shopping sites in the UK, France, Germany, Spain, Italy and Austrian were unavailable Sunday for more than a half-hour, various press reports say, or until around 2145 GMT.

Less well-publicized was the fact that four of Amazon's cloud services operated out of its data center in Ireland were impacted for two hours including EC2, SimpleDB, Simple Queue Service (SQS) and Simple Notification Service (SNS).

Naturally suspicions immediately turned to WikiLeaks' Anonymous army of marauding defenders who for the last week have been unleashing distributed denial of service (DDOS) attacks on the web sites of any and all institutions that have refused to give aid and comfort to the rogue site. And, of course Amazon kicked WikiLeaks off its cloud.

Amazon, whose infrastructure successfully repelled boarders last week, claims its downtime Sunday was the result of hardware failures.

According to the Amazon Web Services Health Dashboard, two of its network devices went down one after the other.

It says, "Between 12:37 PM and 2:32 PM PST [that would be between 2037 GMT and 2232 GMT] the EC2 APIs in the EU-West region experienced significantly increased error rates rendering the EU APIs unusable for most customers. Running instances were not affected, but EU EC2 users were unable to launch new instances or describe their resources in the EU-West region. These errors were caused by the malfunction of a second network device carrying traffic that had recently been shifted off of another failed network device. The first device was nearly repaired but, unfortunately, was still unusable when the second device independently failed. Because full redundancy had not been restored to this part of our network, this second failure resulted in an interruption of connectivity for the EU EC2 API servers. Our networking team was able to re-route traffic to restore connectivity and, shortly afterward, complete repairs on both network devices involved. Though it's somewhat unusual to have two failures like this in this short a period, we will not accept this as a statistical anomaly. We're going to reassess our strategy for keeping spare devices to have better coverage in situations such as these."

From Cloud Computing Journal  http://cloudcomputing.sys-con.com/