AWS:EC2 Introduces Reserved Instances; Massive Potential Cost Reduction

By M. David Peterson
March 11, 2009 | Comments: 6

If there is any one thing that Amazon Web Services is doing a poor job of it's that of slowing down the pace at which they inject game changing features into their web services ecosystem, therefore freeing up the time I would otherwise be spending writing about these new features.

It could just be me, but something tells me NoOne@AWS plans to take pity on my lack of available free time anytime soon which, to be quite honest, is perfectly fine by me!

So what is it that's got me all in a tizzy this go round, throwing around my "free time" like I just won the "Free Time Lottery"? How about reducing my EC2 bill each month by a potential of 33 to 50% for starters? Or providing a service that guarantees me access to hardware resources when I need them as opposed to when resources equal to a desired instance type become available? Both?

In fact, guaranteed resources and massive potential cost reduction is exactly what Amazon has done with the release of EC2 Reserved Instances. From the recent announcement listed under News & Events on

Amazon EC2 Introduces Reserved Instances

We are excited today to introduce Reserved Instances, an additional pricing option for Amazon EC2 that gives you an option to make a low, one-time payment for an instance to reserve capacity and further reduce hourly usage charges. Reserved Instances are complementary to existing Amazon EC2 On-Demand Instances and give businesses even more flexibility to reduce computing costs. As with On-Demand Instances, you will still pay only for the compute capacity that you actually consume, and if you do not use an instance, you will not pay usage charges for it.

To learn more about Reserved Instances, go to

So what makes this such a game changing announcement? Well, let me answer that question with another question:

What's the best way to jump start a stalling economy?

Of course there's differing opinions from people who know far more about this subject matter than I do. But one way is to provide opportunity for people to spend more money by reducing costs for goods they're already paying for, freeing up capital to be invested into places they otherwise would not be invested into.

Enter Amazon Web Services and the introduction of EC2 Reserved Instances.

While Amazon isn't going to single-handedly bring this economy out of its current recession with the release of Reserved EC2 Instances, they're certainly making every effort to do their part by finding ever creative ways to reduce their own costs to then pass along their savings directly to their customers instead of hoarding up that cash and reporting ever impressive quarterly and annual profits as a result.

That's not to suggest ever impressive profits are a bad thing. But putting more money back into the hands of more people is far more important to our collective economy at the moment than is Jeff Bezos ability to expand his ever impressive personal space program by purchasing even more of Texas than he already owns for building and launching space craft.

Don't get me wrong: I've got nothing against expanding our horizons by pushing the boundaries of space exploration. But at the moment we need more money being spent by Texans (as well as all the other states in the union and regions abroad) rather than for Texas. Of course, generally speaking, what comes around goes around, something that tends to translate into an increasing cash flow finding its way back into any given companies financial ecosystem. In this case Amazon will more than likely see an increased growth rate of EC2 instances being consumed by any given customer and/or expanding their usage of other AWS products such as SimpleDB, or SQS. All good things, /especially/ at a time when survival is a /much/ more common word being used on Wall Street than is growth.

Keep the good growth coming, Amazon!

You might also be interested in:


I did the numbers on the EC2 reserved instances and it seems you need to use at least 4643 hours (about 53.0% uptime) to break even on a one year agreement and at least 7143 hours (about 27.2% uptime) to break even on a three year agreement. If you're not going to exceed those hours of usage then it's not worth it.

I get different numbers: if I want to use EC2 to host a website, I'll hopefully get a least one hit every hour. So I'm still paying $0.03 per hour, which is $262.80/year, plus $325 to reserve a small instance for a year. So I've only saved $15.20 of the $0.10/hour pricing. Not much of an improvement.

@Lance: Can you please explain how you came up with $15.20 savings? No matter how I multiply and divide the numbers, I can't seem to find any logical unit of time in which this number would fit. e.g. per day, week, month, or year. Just to be sure we're working from the same set of numbers, an m1.small instance will run you either $0.10 USD/hr for on-demand or $0.03 USD/hr for reserved. To reserve an m1.small instance for a period of 1 year will run you $325. Using the formula:

(((instance cost per hour) * 24 hours) * 365 days)

... to equate the total costs incurred for both instance-types running non-stop for the period of 1 year you get the following results:

On-demand: ((($0.10) * 24) * 365) = $876.00
Reserved: ((($0.03) * 24) * 365) = $262.80

At this point I think we're both on the same page, as your calculations for the hourly costs for a reserved instance accumulated over the course of a year are the same.

In fact, we're both on the same page all the way through the cost of reserving an instance for the period of 1 year which is $325. Here's where I'm getting confused, though. When you add $262.80 to $325 you get $587.80 which when subtracted from $876 leaves a yearly savings of $288.20.

Using a savings of $288.20/year as our basis, when divided by 12 months we see a savings of ~$24 USD/month. When divided by 52 weeks we see a savings of ~$5.54 USD/week. And when divided by 365 days we see a savings of ~$0.79/day which in and of itself doesn't really sound like much /until/ you factor into things the fact that an m1.small on demand instance only runs you $2.40 USD/day which, as Amazon has correctly claimed, provides a savings of ~33%.

Am I missing something more obvious?

@Lance, @Ouroborus
You do not need your 'reserved instance' to be running to provide value. Just the fact that you have x number of servers (physical or otherwise) immeadiately available will mitigate risk involved in load balancing, utilisation shifts (Quarter end / month end / week end etc), System Recovery and even Disaster Recovery.

BTW, monthly charges for a given non reserved EC2 instance that is running 24x7 will vary from month to month, except over the Christmas / New year and between July and August. Determining why this is so is left to the reader....


>> BTW, monthly charges for a given non reserved EC2 instance that is running 24x7 will vary from month to month, except over the Christmas / New year and between July and August. Determining why this is so is left to the reader....

I may have written the piece, but can I still qualify as a reader?

If yes > December 31st, January 31st, July 31st and August 31st. Why these dates hold significance is a task I'll leave to anyone still paying attention to this thread ;-)

If you choose to reserve a small instance for the three year period ($500 one time fee) it works out to an overall savings of slightly more than 50% per year if your instance is on 24hrs/day.

With Three Year Contract:
((0.03*24)*365)+($500/3) = $429.46 per year

Without Three Year Contract:
(0.10*24)*365 = $876.00

That is a yearly savings of: 446.54 or 50.9%

It does require you to lock in for 3 years, but if you have a good reason to, you can save significantly with the reserved instance option.

News Topics

Recommended for You

Got a Question?