Doran Robinson previously worked for healthcare information technology vendor athenahealth.
Only the U.S. site was affected by the failure. The company says it knows what caused the outage, but is not saying.
The U.S. web site of Amazon.com Inc. was out of service for about two hours today, beginning around 10:25 a.m. Pacific time, a spokeswoman confirms. International sites were not affected, she says.
The retailer, No. 1 in the Internet Retailer Top 500 Guide, knows what caused the outage, but isn’t saying. “When the site experiences problems, on those rare occasions when it does, we just work to get it up as quickly as possible,” she says.
The problem likely was caused by a change in Amazon’s web infrastructure, says Shawn White, director of external operations at web and mobile test and measurement company Keynote Systems Inc., which monitors Amazon.com and other web sites through some 3,000 computers around the world. Keynote’s network detected the problem at Amazon.com beginning at 10:16 a.m. Pacific time, White says.
The other possible cause would be a surge in traffic, caused by a sudden rush of legitimate customers or a malicious denial-of-service attack in which hackers try to crash a site by flooding it with requests. But in those cases, White says, sites typically slow down and then go offline. “We didn’t see that here,” he says. “The site was humming along normally, downloading in less than six seconds, then exactly at 10:16 it went down. That would indicate something happened, likely someone made a change that would affect their infrastructure.”
White says it does surprise him that Amazon would make a change in the middle of the day, when traffic likely would be high, instead of at night. “Normally, when you prepare to make a change, you plan it out in advance, give notification and make the change in nonpeak times,” he says.
For the first hour or so after the site went down, visitors to Amazon.com saw a cryptic technical message saying an http 1.1 service was unavailable, White says. “That would have been completely meaningless to 99% of the people who would shop at Amazon,” he says. About 11:45 a.m. Pacific time, White says, Amazon put up a screen that said the site was experiencing performance problems and encouraging shoppers who had made a purchase to confirm the purchase was completed.
Amazon.com went down for more than a half hour in August 2006 and for about 45 minutes in June 2005. The last major glitch at Amazon.com that drew attention was the day after Thanksgiving in 2006 when the site slowed noticeably for about 15 minutes after the retailer offered a discount on the popular Xbox video game console.
Some press reports noted that the outage not only caused a fall in the stock price of Amazon, but also of Akamai Technologies Inc., a provider of content delivery services for Amazon and other web sites. An Akamai spokesman says Amazon’s outage was not related to Akamai. “Our networks have been running fine today across all our customers,” he says.