AWS: Can anyone catch Amazon's cloud?

Grazed from ZDNet.  Author: Jack Clark.


One of Amazon's secret weapons is its attitude to profit margins, where it is happy to undercut its cloud computing competitors.

This combines with its economies of scale to make life very difficult for any of its potential competitors, be they small and nimble or large and rich.

When Amazon adds capacity, it has to buy hard drives by the shipping crate and servers by the rackload, according to Carl Brooks, an infrastructure and cloud computing analyst at Tier 1 research. And the low prices that Amazon can command for its equipment due to its size could make it difficult for competitors unless they are well funded and content with entering into a price war.

"[Amazon's] margins for selling compute and storage are absurd," Brooks says. He notes that a basic EC2 unit — the workhorse of the AWS cloud — is around one half of one modern CPU (a single EC2 unit is roughly equivalent to a 1.0-1.2GHz AMD Opteron or Intel Xeon processor from 2007, according to Amazon)...

Brooks points out that a typical bottom line server costs around $1,000 (£645). This server will have four sockets, each backing a quad-core CPU, allowing it to support 32 AWS VMs.

AWS graphic

Amazon has continually added new facets to its AWS cloud service since its launch in 2006. Data: Amazon; Image credit: Charles McLellan/ZDNet

Amazon charges between $70 and $80 per month for these VMs, each of which has an operating price of $2-$4 per month, Brooks says.

"So unless Amazon is doing something incredibly wrong, we know that the cost of what they're running is incredibly low compared to the price they're charging for it," he said.

We know that the cost of what they're running is incredibly low compared to the price they're charging for it.

– Carl Brooks, Tier 1

This is still a much lower margin than that enjoyed by major enterprise vendors, who sell potted private clouds like the virtual computing environment integrated vBlock appliance or the NetApp-based FlexPod for "an absurd amount of money", Brooks says, pointing to the fact a vBlock will retail for around $1m.

Thus Amazon's size, combined with its commitment to low prices, effectively shuts off the major infrastructure-as-a-service market to small companies, unless they want to offer substantially cheaper services. However, these companies don't have a software ecosystem equivalent to Amazon's, which was initially developed to drive the company's main retail business.

Instead, technology-oriented companies have based themselves on top of Amazon by building technically advanced platform-as-a-service products and layering them over its basic IaaS services. However, though these companies — among them Heroku and EngineYard — have all been successful, they are like the pilot fish swimming by the shark: their existence depends on Amazon retaining its dominance of the IaaS market.

Deep pockets

The only credible competition seems to be from other major IT companies with deep pockets, such as Google, Microsoft and HP.

To that end, Microsoft has recently announced plans to launch new Windows Azure services for rolling out hybrid clouds, including those running Linux virtual machines. This has seen the software specialist move its platform-as-a-service technology down the stack to take on Amazon in infrastructure-as-a-service.

Google is rumoured to be planning the same approach with its Google App Engine product, though specific dates aren't known. Furthermore, Google has poor form in this area — a Google-based cloud storage service was rumoured to be in development in 2006, the year Amazon launched S3, but Google Drive only appeared this year. Even the world's biggest search company, it seems, finds the technology of cloud computing difficult.

Both Google and Microsoft will need to invest a lot of money in the technology, and be able to prove they can run it at scale. Like Amazon, both companies are not immune to failures — Azure went belly up earlier this year, and Google's Docs and Gmail services occasionally fail.

Then there is HP, a company beset by problems, mired in confusion and with a demotivated workforce facing uncertain employment prospects. However, HP is also a company that has proved, time and time again, that it 'gets' technology research and development, so in time HP's OpenStack-based cloud could be a possible contender, if the company can demonstrate a clear vision and the will to make it so.

Amazon's cloud seems to enjoy a number of advantages that come from its scale, ecosystem and relationships with smaller companies that make it hard to beat.

If major multinationals choose to compete with it head on, as we expect they will, the fight will be a long one and could see businesses caught in a race to the bottom on price. Though this would benefit consumers, it could cause more chaos in the cloud.