Amazon Still Resolving Problems on Second Day of Outage

Amazon.com reported that problems that shut down some of its Web services—namely its Elastic Compute Cloud, Relational Database Service and Elastic Beanstalk—still hadn't been completely resolved by late afternoon April 22, more than 40 hours after they went offline.

The most critical outage started at 1:41 a.m. PDT April 21 at an AWS (Amazon Web Services) data center in Northern Virginia and caused disruptions in its EC2 (Elastic Compute Cloud) hosting service, knocking thousands of Websites—including such popular ones as Foursquare, Reddit, Quora and Hootsuite—off the Internet.

These Websites and numerous smaller sites were still offline in some systems by late afternoon April 22. Businesses that depend on the AWS hosting service lost money during that window of time—income that cannot be regained.

The AWS Elastic Beanstalk, which software developers use for deploying and managing applications in the AWS cloud, was running but was experiencing performance problems, Amazon said. Elastic Beanstalk automatically handles the deployment details of capacity provisioning, load balancing, auto-scaling and application health monitoring.

Amazon reported April 22 on its status Website that it has made progress in fixing the outage.

"We continue to see progress in recovering volumes, and have heard many additional customers confirm that they're recovering. Our current estimate is that the majority of volumes will be recovered over the next 5 to 6 hours," Amazon said Friday morning. "As we mentioned in our last post, a smaller number of volumes will require a more time-consuming process to recover, and we anticipate that those will take longer to recover."

By late afternoon Pacific Time on Friday, Amazon had issued a total of 19 updates on its status page since the outages began. EC2, Relational Database Service and Elastic Beanstalk were still having problems.

SLA Not Violated, However

Lydia Leong of Gartner Research wrote in an advisory that Amazon EC2 didn't actually violate its service-level agreement when the outage occurred.

"Amazon’s SLA for EC2 is 99.95 percent for multi-AZ deployments," Leong wrote. "That means that you should expect that you can have about 4.5 hours of total region downtime each year without Amazon violating its SLA.

"Note, by the way, that this outage does not actually violate their SLA. Their SLA defines unavailability as a lack of external connectivity to EC2 instances, coupled with the inability to provision working instances. In this case, EC2 was just fine by that definition. It was Elastic Block Store [EBS] and Relational Database Service [RDS] which weren't, and neither of those services have SLAs."

An Amazon spokeswoman didn't respond to an eWEEK query by end of business April 22.

Mixed Reaction from Customers

Reaction to the outage from cloud customers was mixed.

"Proponents of cloud computing aren't going to like the fact that Amazon had issues that resulted in outages among its customers' sites, but the fact is that most insurers have their own outages when they host applications internally, in some cases with more frequency and severity than we're seeing here with Amazon," Craig Weber, a senior vice president of the Insurance Group at Celent, a Boston-based financial research and consulting firm.

"This outage should focus the discussion on the relative reliability of various approaches and the tradeoffs between them. Of course, there are also lessons about being aware of the capabilities of your business partners.

"Engaging with an SAAS [software as a service] vendor requires understanding things like their architecture, their disaster-recovery capability and similar issues, because worst-case scenarios always seem to emerge eventually."

Morphlabs, which was one of the first AWS solution providers when it launched Morph Appspace in 2007, now has more than 4,000 users. Founder and CEO Winston Damarillo told eWEEK that organizations need to focus on two things to help get a better understanding of their cloud solutions: diversity and control.

"A multi-vendor approach to the cloud means that an organization is not relying on one company or solution to keep its cloud in working order," Damarillo wrote.

"With the hybrid cloud model, companies are able to extend existing infrastructure resources without isolating themselves. When the time comes that they have maxed out their hardware compute capabilities behind a firewall, they can easily make use of the public cloud, as well."