For a few hours over the weekend, thousands of vintage-looking photos went unshared, online movies went unstreamed, and digital scrapbooks went unpinned.
An electrical storm that swept across the East Coast late Friday knocked out power at a Virginia data center run by Amazon Web Services. The outage disrupted numerous web companies that rely on Amazon's virtual data servers, including Instagram, Netflix and Pinterest.
Amazon said a "power outage caused storage failures” for customers who rent space on Amazon’s servers through its cloud service. It said it had restored service to the remainder of its customers on Saturday.
"While no amount of downtime is acceptable, in the six years of running these services our customers have been quite pleased with our operational performance," Amazon spokeswoman Tera Randle said in a statement.
Analysts said the outage highlighted the tradeoffs of cloud computing, an increasingly popular method of outsourcing computing power and storage to remote servers over the web. A growing number of companies and government agencies are taking advantage of cloud computing largely because it is cheaper -- they only pay for the computing power they need. It also allows them to quickly respond to spikes in traffic by expanding their server capacity.
But the cloud is not always dependable. In April 2011, Amazon’s cloud service was brought down by a technical glitch, disrupting several major websites, including Foursquare, Reddit and HootSuite.
Experts say failures at data centers are not unusual. But a failure of Amazon's cloud service can have widespread impacts across the web because the company has thousands of customers.
“An outage at a data center happens every day, but you never hear about them because there are only one or two companies involved,” said George Reese, chief technology officer for enStratus, which helps companies manage their data in the cloud. “But when the cloud goes down, so many companies are impacted it makes big news.”
Amazon allows companies, for an extra fee, to spread their data across eight regional data centers around the world. This "insulates companies against occasional failure" at a single data center, an Amazon spokeswoman said.
Yet most companies, particularly smaller startups, don't back up their virtual infrastructure in multiple places, said Lydia Leong, an analyst who covers cloud computing for Gartner Research.
"A very large number of companies don't have a disaster recovery plan," she said.
Software developer Benjamin Coe said that his startup, attachments.me, which integrates email attachments with programs like Dropbox, can't automatically switch over to a standby computer server during an Amazon cloud outage because that would double infrastructure costs and make operations more complicated. Coe said he spent a lot of time Friday night helping his company recover from the outage.
"Ultimately it’s a trade off," he wrote in a blog post. "Are the risks associated with parts of a system not having redundancy offset by reduced infrastructure costs and complexity?"
"From my perspective," he added, "a lack of total redundancy can sometimes be an acceptable risk if approached responsibly."
While some startups may be unwilling to invest in a backup plan, others seem ready for an unforeseen disruption. During the 2011 outage, Netflix said its services were not affected because the company had backed up data in multiple Amazon data centers.
"Netflix kind of expects Amazon will go down from time to time, so they don't put their eggs in one basket," said Dave Roberts, a senior vice president at ServiceMesh, which helps companies manage their data in the cloud.
On Monday it remained unclear why Netflix's service was affected by the weekend's cloud failure. A Netflix spokesman told The Huffington Post that the company was still investigating the incident, but declined to comment further.
Analysts said this weekend's outage may cause some companies to reconsider moving their operations to the cloud. But experts said that the benefits of the cloud still outweigh the risks, and that companies should focus on building applications that can respond to future disruptions.
"I think anybody who is characteristically nervous and risk sensitive will always see a cloud outage and get more nervous," Roberts said. "But most people who are savvy about the cloud understand it's not infallible."
CORRECTION: An earlier version of this story indicated that attachments.me does not back up its data. The company backs-up its data but is unable to automatically switch over its operations to a standby computer server during an Amazon cloud outage.