Moore’s Law states that computer system performance/price ratio will double every two years. And that was very much my expectation when GoodData started using Amazon Web Services almost 2 years ago. But I had to wait until today to see Moore’s Law at work: Amazon announced 15% drop of EC2 prices. The price of the small Linux instance was constant at $0.10 per hour for the last two years – now it will be $0.085.
15% in 2 years – not exactly the exponential growth in the performance/price curve that I expected. And I started to wonder why. Here are my two explanations – I believe the second one is more likely:
- AWS prices were set way too low to attract developers two years ago. Moore’s Law helped the price to catch up with the real cost of running the cloud.
- AWS is a monopoly and Moore’s Law does not apply.
What? Cloud and monopoly? Isn’t utility computing a perfect example of fiercely competitive commodity where the price curve is shaped only by demand/supply? What would Nick Carr say? Unfortunately not. As much as we read about different cloud providers, AWS is the only real provider of “infrastructure as a service” in town. If you don’t want to be locked-in to proprietary Python or .Net libraries there is not that much choice.
Until we will see performance/price of AWS double every two years, we should still wonder about monopolistic pricing.
Comments
Moore’s Law actually states that the density of transistors that can be placed on an integrated circuit will double roughly every two years. Clearly this has implications for the performance/price ratio of anything that depends on microprocessor power, but it doesn’t obviously apply directly to the technologies that underpin Amazon’s offer.
I wouldn’t go as far as to call AWS a monopoly, but you’re probably right that lack of competition has something to do with the relatively slow decline in their prices. But I suspect it has more to do with the fact that the current prices reflect a lot of things besides the cost of transistors (system admin costs, software development, bandwidth, etc.) and that these are dropping in price much more slowly.
Moore’s law certainly isn’t the only parameter on which to base such an analysis: what about energy consumption? If power doubles, but energy consumption doubles too (Whic doesn’t, I agree
, your performance/price ratio stays the same.
There are also other maintenance costs. And I wonder how many servers they have and what their average life time is. It would be an interesting analysis to determine the optimal life length of a physical server in the AWS server farms.
There’s also an upfront investment that Amazon would be glad to get back.
And I’m sure there are tons of things that have to be added. But in the end, we certainly shoudl get more than 15% in 2 year.
They are not a monopoly, and Moore’s law stopped working a while ago for computer chips, so it makes sense that it won’t work for cloud computing, which in many ways is a derivative of – wait for it – computer chips!
According to Wikipedia (http://en.wikipedia.org/wiki/Moore%27s_law#Ultimate_limits_of_the_law) Moore’s Law is still going strong and expected to do so for at least another decade.
You’re saying it’s likely the AWS is a monopoly environment on one hand and on the other hand you’re expressing your unwillingness of being “locked-in to proprietary Python or .Net libraries” which I understand as critics of the Google App Engine proprietary implementation mostly in the BigTable part of the stack. What exactly is the difference between deploying the environment of your choice into a monopolised AWS and using, as you say, proprietary Python or Java libraries on Google App Engine when you can’t switch to any alternative provider with ether solution?
Monopoly is measured by the market share and not by the level of openness. Many global telcos are monopolies while they deliver perfectly standard GSM and other open services.
I am not locked in AWS because it is proprietary but because there is no other alternative. Open provider with 100% market share is as bad as proprietary one with 1%. Both can use their market power to manipulate prices.
I was wondering a few weeks ago about the same – if there is any plan of Amazon to push down the price of EC2 more significantly. I tried to Google for “will amazon ec2 decrease prices” and Google replied by: Did you mean: “will amazon ec2 increase prices”? I find it kind of funny
Almost seemed to me like if you Google “french military victories” and click the I’m feeling lucky button (or just following the first result)
If there is no lock in, and Amazon is using its status as the only viable offer to overcharge, wouldn’t cheaper competitors immediately appear? I don’t know what the truth is, but it seems more plausible to me that the lack of competition is precisely because Amazon is charging relatively low prices, leveraging the enormous scale of its server infrastructure.
None of this should prevent us from making fun of the French, of course.
It’s too short time that this cloud services are here. They may just calibrate it.
Trackbacks
One Trackback