I’m not a huge fan of the IaaS term. To me it wastes too many letters (aa adds what value, now?), and makes the phenomena of cloud computing infrastructure the little brother of SaaS. Never the less, you are likely to see it in the opening slide deck of any major cloud computing conference you might attend, as if it has been blessed by the mighty Gartner itself.

One of those middle wasted letters could surely be put to better use, and probably both. IaaS doesn’t tell me enough about the context of the ‘service’ or the scale of the service. We’ve had hosted infrastructure for years, accessible over web interfaces.

Here is what’s so cool about clouds like Rackspace: the applications can now access this infrastructure without me meddling with a human oriented browser interface.

I can write an application in Python that automatically scales itself up by grabbing more servers as load increases. I can also expect that there are enough resources in the pool to get me to at least ‘web scale’ where millions of people can access them at once without the cloud running out of cycles, storage and bandwidth.

Infrastructure clouds are much closer cousins to SOA than to SaaS. They exist for applications to be built on top of their framework and to scale automatically. SaaS and IaaS are not as focused on the upstream composite creations.

Therefore, I’d like to suggest for our next conference slide deck we aren’t as lazy and make the first layer of the stack:

SOSI

Service Oriented Scaled Infrastructure

What this abbreviation tells me:

1) Its web service oriented, can be called in a ‘self service’ (thanks @jamesurquhart ) no human interaction style programmatically over a network interface. Its designed to be a functional layer of a construction, not a terminating point (say hosted MS exchange). This also keeps out the virtualization=cloud #cloudwashing. It isn’t enough to be virtual—those resources must also be easilly called ‘upstack’ in a flexible way.

2) Its at least web-scale in capacity. There are a lot of rants going around about how big grid applications could consume any cloud—correct. HPC can always out demand any system, true. However, web scale has some meaning to me and I like it. It tells me I can build a service to be used by millions to hundreds of millions and the capacity will be there. This also keeps us from the silliness of calling any pool of virtual resources in a data-center a cloud. (#cloudwashing)

Now all I need is a button that says “No Servers, Only SOSI”.