By Joe Masters Emison

Who can beat Amazon at IaaS?

By Joe Masters Emison

Amazon is the clear number one in Infrastructure-as-a-Service offerings.  Depending upon your definition of IaaS, number two is probably Microsoft, although it could be Rackspace–no one is publishing enough information to know for sure.

Amazon is obviously way in front; the most recent evidence of this is that Microsoft Azure had a 60+ hour outage this week, and searching for “Azure Outage” on Google News yields more stories about the Amazon Web Services outage two weeks ago than Azure.

So who’s got the best chance of beating Amazon?  My money is on Google.

Google is the only other company that has

  1. enough money and equipment to compete on price, and
  2. enough technical expertise to pull off what AWS has done.

And that may sell Google short: the core of AWS is just a re-implementation of what Google invented in the first place. Several recent pieces in Wired cover this: Amazon sneaking peeks at Google servers, and Inside a Google Data Center.

Also, Google has one killer feature that AWS does not have and likely will not build any time soon: a global private network.  This means that regions for Google will be much more connected than AWS regions. 

For example, you cannot share or transfer AMIs or EBS snapshots across AWS regions, but you should be able to with Google Compute Engine.  This immediately makes GCE more attractive for greenfield deployments that require high availability and/or have low RTOs.  And there really isn’t a price difference (or GCE is cheaper–I haven’t seen or run enough benchmarks to have a definitive answer here).

What about Microsoft? Microsoft has data centers around the world, and a more widely-used (and more unique) IaaS offering than Google. However, Microsoft’s Linux marketing has been lackluster, and with so much of the IaaS usage (and growth) around Linux, I don’t see Microsoft getting to number two before Google does.

In particular, I see the internal conflict of interest at Microsoft (promote Windows vs. have full OS support within Azure) as somewhat similar to the ones that plagued Sony around digital music (wanting to make good MP3 players vs. DRM requirements from Sony Music). 

Azure is not going to have anywhere near the margins of Windows, so the marketing efforts of Azure simply can’t go head-to-head with AWS or GCE on Linux without someone freaking out about cannibalization.

I certainly expect–at least in the short term–for Microsoft to remain successful in selling solutions (Cloud included) to the same people who have been buying Microsoft for years.  But if a provider stands a chance of beating Amazon, I think it will need to capture a lot of the Linux market. 


While it’s theoretically possible to orchestrate/automate Windows servers as one can Linux servers (with something like Puppet), it is much harder and less native/intuitive to do so with Windows.  The Windows standard practices revolve around humans using GUIs; Linux/Unix standard practices (command line, text configurations, one-function-per-tool) are much easier to automate, and fit much more naturally with IaaS architectures than Windows.

So it’s Google. 

Good luck, Google!  We need the competition, if only to keep the prices low.

  • Craig

    Google won’t get into IAAS. And Microsoft is PAAS with IAAS just an afterthought for legacy applications. What I expect in the long term is Amazon to get into PAAS as that it the most logical solution for the majority of cloud services outside of the top 1% who need a lot of customisation.

    • “Google won’t get into IaaS” — ? What do you call Google Compute Engine? (https://cloud.google.com/products/compute-engine) Here’s what Wikipedia says about it (http://en.wikipedia.org/wiki/Google_Compute_Engine):

      “Google Compute Engine (GCE) is an IaaS product announced by Google at Google IO on June 29, 2012.”

      (I also disagree with all of the other statements in your comment: Azure is being used–and is heavily invested in by Microsoft–as a top IaaS; Amazon’s PaaS solution is services like CloudFormation; far more than the top 1% will use IaaS for a variety of reasons, not least of which are BC/DR and regulatory requirements).

  • Joshua Christensen

    The Azure ‘outage’ was >6 hours and it mostly affected their beta products – which they recommend that you do not use for production – or ‘degraded’ (note, not ‘outage’) of their other services.

    Microsoft has a huge advantage Google’s no-where-near-ready solution: It’s not limited to MS technology. Azure is extremely open and if doesn’t support what you need, you can create a virtual machine and throw whatever you want on it. Google is only running Linux virtual machines. Which if it’s only VMs you’re going to need someone to administer them. At least with Azure you can get everything running in 5 minutes or less.

    Amazon’s going to always be solid unless something really bad happens with them and they lose everyone’s trust (which is highly unlikely). One big advantage Amazon has is that they too are very open about what you develop with. However, the .NET community is very large and the integration that MS can get with it’s products is by far superior to what Amazon can do.

    I honestly don’t think MS is aiming at getting the cloud business of people not on the .NET platform. I’d be willing the bet they’re looking to keep it all “in house” and create a very tightly integrated solution around their technologies. But also realizing that not everything may be within the control of moving over to the Azure platform and thus offer support for anything else.

    • We’ll have to agree to disagree over whether Windows VMs are easier to administer than Linux VMs. I pay our Windows sysadmin for administering 5 Windows servers the same as I pay our Linux sysadmin for administering 35 Linux servers (that flexes hundreds of servers when loads go up). There are so many more tools and resources for orchestrating Linux than Windows. Ultimately, orchestrating Windows is about having to navigate Powershell, which is essentially an imitation of the Unix commandline–a more poorly-implemented, less-well-documented, and much-smaller-community-supported commandline than Unix/Linux’s commandline operations.

      But in the end, that’s really the only thing we disagree on. I made the same points about Microsoft not targeting outside of the Microsoft world above–you just happen to disagree with me about the prevalence of Linux in the IT world. I’ll just call “Scoreboard” to support my side: how many of the top 100 websites/web applications in the world run Windows? What’s the Windows vs. Linux market share breakdown for web sites / web applications?

      After all, IaaS is primarily used (and will likely continue to be primarily used) for hosting web sites / web applications. I don’t see .NET making significant movements there, and I would expect Linux to continue its dominance. Hence, Microsoft is not set up to beat Google in the IaaS race, even if it takes Google a year or two to get out of beta for GCE.

  • Ahmad

    I think Softlayer is very promising. Take a look at softlayer.com

Get the latest in Entrepreneur, once a week, for free.