Simon’s Blog

Cloud computing is currently one of the hottest areas in technology.  In recent days, we’ve seen new cloud computing announcements from both Amazon (with upgrades to its EC2 computing platform) and Rackspace (who announced some acquisitions, and a whole new offering going forward).  Google is in the market already with its App Engine platform. Today, Microsoft announced its cloud computing offering, called Windows Azure.  Next year, Sun Microsystems will announce its cloud computing plans codenamed “Project Hydrazine” (assuming the problems the company has with its share price don’t lead to some kind of event that changes their plans).   Cloud computing is an interesting value proposition for developers, and they’re are starting to use parts of these services in significant numbers.   The big question is: do any cloud computing vendors have an offering that provides a complete solution for enabling people to build scalable web businesses?

The Test

Here’s what I think the test is of whether a cloud computing offering provides a complete solution. It needs to provide:

• All the hardware and software infrastructure required to serve up complex web-sites with fast load times  (say less than two hundred milliseconds load time for a simple page)
• The ability to add and remove server capacity automatically and dynamically as demand rises and falls (and of course to pay only for what you use)
• To be highly available (think better than 99.5% availability)
• And… critically, to do so in a way that is much cheaper for people than buying and running their own hardware

Or, to put it more simply: would it be both possible and cost-effective to build the next Google on a given cloud computing platform, without running any hardware yourself?

If that’s the test, then – do any of the current cloud computing vendors pass?  As far as I can see, they don’t.  How do I know this?  Simple. Sizeable Internet companies such as Digg and Facebook; and even small, ambitious start-ups such as FriendFeeed and Mahalo haven’t been rushing to move/build their entire infrastructures to/using cloud vendors.  Why?  For two reasons: firstly, from a technical point of view, the cloud vendors simply can’t meet their needs; secondly, even if they could, using cloud vendors would be much more expensive than running their own hardware. To be fair, Microsoft’s offering is too new to evaluate, and we don’t know what Sun’s offering will look like yet, so the situation may yet change.

Is There An Elephant In The Room Cloud?

I’m pretty sure that cloud computing vendors would want to offer all of the above if they could.     Today, however, cloud computing services offer neither the required performance, nor competitive prices for running a complete web business infrastructure.  The absolute best-case is they offer a good deal for some parts of operations running at particular levels of scale.

This, then, begs the question: is there an elephant in the room?  That is, is the truth that there actually aren’t any significant general economies of scale in building really large virtualized compute offerings?  In other words, are we witnessing a huge amount of hype around what will actually turn out, in the end, to be less than a zero sum game?  I don’t know the answer.  What I do know is that the cloud computing vendors seem to be some away from being where I think they need to be. They can’t even seem to get the price of RAM competitive, and that’s just a matter of buying sticks of RAM and putting them in a box. If I’m wrong on that, by the way, please point me to a cloud computing vendor that offers virtual servers with 32 or 64 GB RAM at a reasonable price; “reasonable” being, over a three year period, much less than the cost of buying the RAM outright (actually, forget about cheaper, getting close to the price would be a good start).