Amazon Web Service Outage Shows Need for Decentralized Cloud Computing -

Amazon Web Service Outage Shows Need for Decentralized Cloud Computing


Editor’s note: This article has been updated to fix a typo. The original text stated that there was a rise in errors for Amazon S2 requests. The corrected text states that it was Amazon S3. 

This morning, Amazon Web Services (AWS) underwent a major cloud outage, causing both its Elastic Compute Cloud and Simple Storage Service solutions to receive many request errors in the United States. AWS is a cloud computing solutions provider ran by Amazon, a company most known for its large online retail network

Also read: Amazon’s Approach to Comic Book Distribution Resembles Bitcoin Ideology

08bits-amazon-tmagArticleAmazon began investigating a rise in errors for requests made to Amazon S3 in the US-STANDARD region at 12:36 AM PDT. Just a few minutes after this investigation began, the company reported even more errors for the EC2 APIs and launch failures for EC2 instances in Amazon’s US-EAST-1 region. Additionally, AWS experienced heightened latencies in its SendEmail, SendRawEmail, and Send SMTPEmail APIs in the US-EAST-1 region.

Two AWS customers, Heroku and GitHub, have confirmed that they have experienced issues due to the AWS outage. Heroku reported that customers could not push changes to their apps, while GitHub users saw increased error rates on repository release downloads and issue image uploads.

Amazon reported at 1:52 AM PDT that:

“We are actively working on the recovery process, focusing on multiple steps in parallel. While we are in recovery, customers will continue to see elevated error rate and latencies.”

Finally, at 3:46 AM PDT, Amazon notified customers that its AWS errors had been corrected and services had returned to normal operations.

AWS Outage Highlights Need for Decentralized Alternatives

The root of the problem with this morning’s outage lies in the fact that AWS is a centralized cloud computing service, meaning it has a central point of failure. Problems in one area of AWS’ services could lead to problems for the rest of its customers, as Amazon’s servers and support services are all centralized. Using some kind of decentralized network — such as a blockchain-based alternative — would virtually eliminate the possibility of major service outages like the one experienced by AWS customers this morning. With a blockchain alternative to centralized cloud computing, one node may go down, but there would be countless others to pick up the slack, each one capable of recalling data from the blockchain. With this method of cloud computing, chances that users would experience any significant amount of cloud downtime are extremely low.

An example of  decentralized, blockchain-based cloud computing is Storj, a decentralized cloud storage service. Storj is secured by a peer-to-peer network powered by the blockchain, which ensures that stored data is never lost and is always available. Certainly, as cloud computing becomes more of a necessity in our modern world, additional problems caused by centralized services will lead people to migrate towards decentralized alternatives similar to Storj.

Do you think a decentralized cloud computing solution would prevent outages like the one AWS experienced? Let us know in the comments below!

Sources: InformationAge, TechWeek Europe

Images courtesy of Logi Analytics, The New York Times

Evan Faggart

Evan Faggart

Evan is the Senior Editor of He has a bachelor's degree in History with minors in Economics and Political Science. When he's not acting like he knows what he's doing in the newsroom, Evan is most likely playing video games. Follow Evan on Twitter @EvanFaggart.

  • Joshua Wiens

    You are kidding me with this right? Their uptime is insane, and your are throwing a fit because they had an issue.

    Someone was looking for a reason to spin it into a ‘decentralization is the solution’ article.

  • Genesis R. Miranda

    People who doesn’t understand AWS should not talk about AWS. FYI AWS is NOT centralized (depending on implementation) It has multiple regions and multiple availability zone per regions, administrator just need to spread out their instances and other Amazon Web Service across regions to decentralize their infrastructure. And still the uptime of each region is very impressive.

  • CoinHedge

    Yes it is. More decentralization is good. By the way i mend some DigitalNote XDN on Amazon, is very good for CPU mining

  • Neal

    Are these articles proofread? There isn’t an “S2″ service in AWS.

    As to the larger point of the article, it might help to make the finer distinction that AWS is both distributed and not — depending upon the service you are talking about. Services like Route53, CloudFront, S3, are basically “global” services, while EC2 and others are region-based.

    What confounds me is that the lead paragraph of your story explains the US-East-1 regional scope of the issue for EC2 and some other EC2-based service issues in that same region, but then your analysis section declares a need for AWS to be decentralized. This is a contradiction, since talking about US-East-1 is talking about one of the decentralized segments of AWS. With 9 global regions for EC2 to live in, and many more availability zones within those regions (and each AZ represents > 1 data center), I’m baffled by your analysis.

    Perhaps a more helpful and interesting angle for a report would be how companies using AWS and other providers still fail to design against that architecture for HA, either assuming that the “cloud” cannot fail them, or that putting all eggs in one proverbial basket is still good enough. But the take-away from this article is fundamentally wrong.

  • Evan

    Hi Neal,

    Thanks for your feed back. “S2″ was a simple typo, and we have fixed the article so it reads “S3.” Your comments on the rest of the article are interesting, although we do not believe the analysis is “fundamentally wrong.” Of course, a thorough analysis on AWS and whether or not the blockchain could improve it would take more than a few hundred words, and we believe that these issues would make a good full-length follow up to this article.