Nature's clouds prove more powerful than Amazon's

As a major storm crushed the Virginia and Washington, DC area last night and disrupted power to over two million customers, the thing many millions more around the world noticed was that they could no longer share pictures on Instagram, could no longer post interesting tidbits on Pinterest, and couldn't stream a Netflix movie Friday night after dinner. These three sites, and many more, use Amazon's "elastic cloud."

The service is supposed to have so much redundancy built in that a failure in one data center simply routes traffic to another data center. Unfortunately, just like last April, this outage was noticed worldwide. Even users in Europe, who have their own Amazon hubs nearby, were impacted by the outage in Amazon's north Virginia data center power outage. This is significant because Amazon's stated service level agreement (SLA) for EC2 is 99.95% - which means they promise only four and a half hours of downtime a year. This outage alone has reduced the uptime to less than 99.95, and that's not taking into account the outage Amazon suffered in the same area just two weeks ago.

With everyone clamoring to get into the "always-on" public cloud, incidents like this have to make people take a step back and think about the ramifications. The allure of the cloud is that your data is everywhere for everyone, but as history has shown, things aren't always as smooth as companies would have us believe. Incidents like this can have major financial impact, especially as more and more companies are moving their services into the public cloud arena.

Via: Amazon's AWS Status Page

Report a problem with article
Previous Story

Leak: Office 2013 icons revealed

Next Story

Microsoft's OEM head departs; unrelated to Surface reveal

31 Comments

Commenting is disabled on this article.

There's nothing wrong with Cloud Computing, what's wrong is Amazon's incompetence who advertise "redundancy" but don't even test it for even the most basic of faults.

Simon- said,
There's nothing wrong with Cloud Computing, what's wrong is Amazon's incompetence who advertise "redundancy" but don't even test it for even the most basic of faults.

Well said! I would hope that a company as big as Amazon would have IT staff running down lists of what ifs, hell I know I sure would! Also if there was ample enough warning that a storm could be happening, that should give Amazon techs enough time to make sure everything was in working order!

No wonder some sites were telling me Netflix was down and others said the opposite. I also found it weird that I could access the website from the pc but couldn't connect through XBL.

fmanchu said,
Where the hell is "North Virginia"?

Good question. Amazon refers to their data center location as "north Virginia" and "north California," even though I don't actually see a "south Virginia" or "south California" listed. Curiously, they don't do that for Oregon...

xendrome said,
I'm pretty sure this was a connectivity issue not just related to Amazon but that whole region..

But the point is that the elastic cloud is supposed to re-route traffic to other locations in the event of a data center outage, and that didn't happen.

xendrome said,
I'm pretty sure this was a connectivity issue not just related to Amazon but that whole region..

they said it was power

amazon said,
We can confirm that a large number of instances in a single Availability Zone have lost power due to electrical storms in the area

Doesnt make sense. I remember something about Amazon saying they had redundancy to where data and whatnot was stored in 3 separate geographical locations. Service should not have been affected.

Beyond Godlike said,
Doesnt make sense. I remember something about Amazon saying they had redundancy to where data and whatnot was stored in 3 separate geographical locations. Service should not have been affected.

Unless they were talking out their bum holes

Beyond Godlike said,
Doesnt make sense. I remember something about Amazon saying they had redundancy to where data and whatnot was stored in 3 separate geographical locations. Service should not have been affected.

sure, but this is a switching station of sorts not a data backup. That said they probably should have a second one at least in case stuff like this happens

What sort of a datacentre loses power? Haven't they heard of UPS's and backup generators? Where is this Amazon cloud hosted? In someone's garage or bedroom in Silicon Valley??

dvb2000 said,
What sort of a datacentre loses power? Haven't they heard of UPS's and backup generators? Where is this Amazon cloud hosted? In someone's garage or bedroom in Silicon Valley??

I was affected as well.... and gas was hard to come by even diesel was. I know datacenters have fuelling contracts, however they might not have enough fuel for the generators for a whole week.... most have it on hand for 2 days. The lines to get gas were sometimes 6 hours.