Cloud Watch: Amazon Explains Its OutageCloud Watch: Amazon Explains Its Outage

On Friday, Amazon explained the pre-Easter cloud outage that took down several Web services -- not just Foursquare, but things that are actually useful, too.

It seems the problem was a configuration error that made Amazon's machines think an enormous number of disk drives had failed. A big recovery job started, paralyzing Amazon's Elastic Block Storage. On All Things D, Arik Hesseldahl described it as "a human-caused mistake that was then exacerbated by the way the cloud system was designed to work."

There were ways around the Amazon pileup. Charles Babcock at Information Week talked to one company that redirected all its traffic to Amazon's West Coast facilities.

Customers got a 10-day credit and an apology. Babcock mentions hasn't heard from anyone who's thinking of switching services.

And The Economist mentions the case of Netflix Inc. (Nasdaq: NFLX), which apparently uses the Amazon data center in question but didn't have any problems. Netflix does what you'd think would be obvious: It designs parts of its infrastructure independent of one another, so they don't all go down at once. Netflix calls this the Rambo Architecture, possibly in tribute to the disjoint segments making up Rambo movies.

In non-Amazon news, evidence of an Apple Inc. (Nasdaq: AAPL) cloud service continues to build. After reports broke that Apple bought the domain name iCloud.com, French site Consomac.fr found evidence of a cloud feature in Leo, the upcoming 10.7 version of OS X. It's code-named Castle, as Consomac posted over the weekend.

The going theory is that Apple will provide cloud storage for photos, videos and music, the last one being done with the cooperation of Warner Music. Its 500,000-square-foot data center in North Carolina, being built to support Apple's iTunes and MobileMe services, is the likely home for Castle, too. (See Apple Prepping Cloud-Based Video Service?)

iCloud.com, by the way, still goes to a site owned by Swedish company Xcerion -- which, according to GigaOm, got $4.5 million for the domain name. Admit it, you checked that already (or wanted to).

It's been a busy few weeks for cloud (in all its meanings):

— Craig Matsumoto, West Coast Editor, Light Reading

Interested in learning more on this topic? Then come to the Carrier Cloud Forum, the latest event in Light Reading's Service Provider IT (SPIT) series. Collocated at Interop in Las Vegas, May 9, 2011, the workshop is free for attendees meeting our prequalification criteria. For more information, or to register, click here.