Amazon Web Services suffers outage, takes down Vine, Instagram, others with it

Amazon Web Services suffers outage, takes down Vine, Instagram, others with it

Summary: The cloud giant suffered an outage for about an hour on Sunday, showing once again the perils of an outsourced cloud service, as many AWS customers went down with it.

SHARE:
TOPICS: Amazon
12
137752249423365-1
(Image: ZDNet)

Amazon Web Services (AWS), one of the world's largest cloud provider, stumbled over on Sunday for 59 minutes, due to issues with its U.S.-EAST datacenter.

The outage began at about 1 p.m. PT (4 p.m. ET) following connectivity issues in the North Virginia datacenter, which led to elevated API error rates in the region. This led to "degraded experience," according to the AWS status page, which resulted in a "small number of EC2 instances [becoming] unreachable due to packet loss in a single [availability zone]."

Last week, AWS suffered downtime that lasted around 25 minutes. Most websites running on the AWS cloud were unaffected. The biggest casualty of the outage, however, was Amazon.com itself, which rejected customers from accessing its site in the U.S. and Canada.

Other Amazon-owned websites also suffered, including Audible.com, while Netflix continued to power through the problems.

While international sites were unaffected, some crunched the numbers, and estimated that the company could have lost as much as $1,100 in net sales per second. By comparison, Google's five minute outage a week before is said to have cost the search giant more than $545,000.

But the knock-on effect was felt. Users of Vine and Instagram, as well as others — IFTTT, Airbnb, Flipboard, just to name a few — fell at the mercy of its cloud computing parent.

Instagram alerted its users of a fault to its service almost as soon as it occurred:

Vine caught up, about half an hour later:

Amazon is not infallible, neither is Microsoft's Azure or Google's Compute Engine platforms. But on the whole, they all fare relatively well — Amazon generally more than others. In spite of this, one of the reasons why so many sites fell down at the time was because the affected datacenter supplies many high profile sites, including the aforementioned.

The North Virginia U.S.-EAST datacenter, above all other Amazon cloud datacenters, seems to get the worst rap out of them all, yet many companies still offer their services there. 

For Amazon, U.S.-EAST is its oldest and largest, but even when Amazon recommends employing georeducancy strategies — simply rolling out a cloud service or application to multiple geographic instances — it's a mystery why so many technology companies still rely on this one datacenter, beyond the realms of supplying "just" the U.S. east coast. 

Topic: Amazon

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

12 comments
Log in or register to join the discussion
  • US-East is too crowded

    Too many users are using that old center probably b/c of the low price.
    LBiege
  • ROI vs Outage risk

    I'm sure the usual suspects deriding Cloud and those for tearing down private data centers will be all over this in a few hours. But this is still a business decision about risk. How much is the risk of an outage to business vs the Cost/ROI of rolling your own. A long as the equation is tilted toward low cost/high availability/high security solution, that is where the trend is going to keep moving.

    End of discussion
    mpompey0110
    • oh really?

      @mpompey0110, you sed, "end of discussion" Um, okay, I will not discuss this with you. Clearly it would be a waste of time. Have fun in your vacuum of one-way conversation.

      For those who ARE interesting in discussion, I think I've said here before, cloud schmoud. This latest outage example just highlights once again that cloud != perfection. It helps debunk the myth that "if we just put everything in the cloud then everything will be great, cost less, perform better, drive profits, and be really cool". For those contemplating the move to putting apps offsite (aka cloud), a careful cost/benefit analysis should reveal whether it makes sense. Do that instead of jumping blindly on the bandwagon for the sake of following the latest fad.

      This latest example also goes to show that, like Microsoft, Google, T-Mobile and others, even multi-billion dollar companies have to work very hard to achieve 3 nines. I scoff at hiring managers who write job descriptions that insist on infrastructure people acheiving 4 or 5 nines. We mid-size/little guys often can do better than the mega giants on reliability for far less cost.
      moebiusloop
      • It's worse when my clients suffer but it's not my fault

        The really bad aspect of outsourced IT (i.e. cloud) failures is that when it affects the outsourcing comany's clients, the outsourcing company looks bad through no fault of its own - unless you consider outsourcing to be a fault!
        JohnOfStony
  • again

    Live by the Cloud, die by the Cloud.
    Doc.Savage
  • Haven't learned their lesson

    Instagram apparently didn't learn anything from their last major outage. Millions in VC money, yet too cheap to have geographic redundancy. Idiots.
    rsmith187
  • Amazon Major Outages

    The Amazon major outages seems quarterly event to me. I did research on Amazon cloud major outages and found some interesting aspects of challenges Amazon has http://www.slideshare.net/rahultyagi50999/amazon-cloud-major-outages-analysis.
    byrahul@...
  • AWS outage - once burned, twice shy - ShortStack for Facebook switched

    AWS outage - once burned, twice shy - ShortStack for Facebook switched from Heroku on AWS to Clustrix on last outage, survived this one.

    http://www.clustrix.com/blog/bid/312144/AWS-outage-once-burned-twice-shy-ShortStack-for-Facebook-switched-from-Heroku-on-AWS-to-Clustrix-on-last-outage-survived-this-one
    Raj Bains
  • Datacenter in Virginia you say ?

    Now which government organization runs out or Virginia ?
    Alan Smithie
  • This is more on Amazon and not on Cloud Providres per se

    We shouldn't blame the cloud mantra or outsourced IT. While inhouse is good it's just prohibitive and expensive. We wouldn't have all these awesome apps on our phones etc without a way to quickly add hosting infrastructure beyond adding physical dedicated hosting equipment/routers/firewalls in our own managed datacenter. It's just not plausible. I guess back to my original point this is entirely Amazon's fault. Don't blame the little cloud providers like us. Where I work we guarantee up-time. As well we're cheaper and just as robust as Amazon.

    We just wrote a blog about it but I don't want to sound too spammy.
    http://www.codero.com/blog/aws-cloud-flight/
    Craig Campbell
  • Don't Put All Your Eggs (Data) in One Basket (Center)

    The best way to have high availability in the cloud is with a multi-regional (cross-datacenter) distribution solution like GenieDB. This will allow your database to remain available during regional outages. Otherwise, you're putting all your data in one proverbial "basket".
    DCJDMBA
  • pay peanuts.... get....

    People flock to these services as they see price as the most important piece in their decision making criteria. Perhaps , performance , reliability and support will soon overtake cut-price commodity mentality.
    Confrontation internally over price of a 'premium' provider is surely better than confrontation, after it becomes apparent that risks were taken because of a aversion to confrontation to begin with and running a POC on price alone.
    Dublin98