Amazon cloud offers protection from falling meteors
(Virtually) infinite disaster recovery
Now that you can purchase future compute power on its intercontinental cloud, Amazon is pitching the sky-high operation as a safe haven for anyone worried that their earthbound data center may soon be hit by a falling meteor.
Last week, Amazon introduced "reserved instances" on its Elastic Compute Cloud (EC2, for short), letting developers reserve capacity up to three years in advance. Previously, you only paid for compute resources by the hour, as you used them. Now, by paying up-front you can save a little dough - while ensuring that all the compute power you need will be there when you need it. At least in theory.
Plus, Amazon says, its cloud is now ideal for large-scale disaster recovery. "We've learned that some of our customers have needs which aren't addressed by the spot-pricing model," Amazon cloud guru Jeff Barr wrote in a recent blog post.
"Quite a few customers actually told us...they were interested in using EC2 but needed to make sure that we would have a substantial number of instances available to them at any time in order for them to use EC2 in a DR [disaster recovery] scenario. In a scenario like this, you can't simply hope that your facility has sufficient capacity to accommodate your spot needs. You need to secure a firm resource commitment ahead of time."
A nice idea. But as our good friends at Data Center Knowledge have asked: Can Amazon ensure that it will always have enough compute power on hand for everyone banking on EC2 for disaster recovery - even if there's a 911-esque catastrophic event that takes out earthbound data centers left and right?
According to Amazon, the answer is "Yes."
"When we sell a reserved instance to a customer, we really are reserving that instance for the customer and we will not commit it to other purposes when they're not using it," Amazon EC2 general manager Peter De Santis tells us. "That's an easier capacity-management problem than predicting future demand. With a reserved instance, they're giving us the ability to know that that instance will be needed...
"It's like we're selling a number of seats on an airplane - and we're not going to overbook the airplane."
So, if you wanted to duplicate all your code and data on Amazon's cloud and reserve enough compute capacity to mimic your entire data center in the event of a catastrophe, you could do so. According to Amazon.
And this would be significantly less expensive than, say, building a second data center on your own. With reserved instances, once you pay your up-front fee you don't pay additional usage fees until you actually used the space you've reserved.
"In the old world, if you wanted a full hot standby in case your data center got hit by a meteor, you had to go off and provision a lot of hardware, buy it, and put it into another data center with the power and the cooling and the people to take care of it," de Santis says. "You had to make an upfront payment for that second site and pay for the operation of the site for the duration of the time you want it.
"With reserved instances, you pay your upfront fee, but if you never need the capacity, you never incur an additional charge. But you know it's there when you need it - and if you're a smart company, you test it once in a while to make sure the fail-over works."
Of course, de Santis also argues that it makes even more sense to put your data recovery setup as well as your in-production infrastructure on the Amazon cloud. EC2 is served up from two separate geographic locations. And each geographic is split into "availability zones" designed not to fail at the same time.
"These zones are engineered to have few low-failure correlations," de Santis continues. "They're not in separate parts of the country, but they are isolated from one another." So, he says, there's no need for you to even build your first earthbound data center.
All of which begs the question: Just how much capacity is up there on that cloud? But as usual, Amazon is mum. All de Santis will say is that there are some Amazon data centers in the eastern United States and others in somewhere in Europe. (Dublin, we're told by those in the know.)
At the moment, 490,000 developers are registered for the Amazon cloud. And they generally have access to as much capacity as they need - outages aside. "It's hard to argue with their track record," says Thorsten von Eicken, the CTO of RightScale, a California operation that provides management tools for the Amazon cloud.
But there has to be a point where the capacity ends. And you can't help up wonder where that might be. ®
Sponsored: Hyper-scale data management