Container based computing – How is this going to work, really?
Posted by Ken Jamaca on Sun, Feb 01, 2009 @ 01:46 PM
With cloud computing being at the forefront of everyone's mind, I am curious how Container Based Computing will play out in 2009. For those that live under mossy rocks, container based computing is the method of building mini, self contained datacenters inside standard sized shipping containers.
Companies like Rackable Systems, Dell, Sun, and a handful of others are moving quickly and decisively towards Container Based Computing as core product offerings. Customers like Microsoft, Google, Yahoo, and Stanford University are already using containers in their production environments. In fact, it is no secret Microsoft plans on using several hundred containers in their Chicago datacenter. That said, I expect Container Based Computing to get very large, very fast.
I heard recently that Microsoft is considering a full container as a single asset. This change in methodology is striking for a few different reasons. A container can hold several hundred servers, network devices, etc. If Microsoft (or any company) starts considering that as a single asset, what happens when 1, 2, 20, etc of those devices fail? If it is a single asset, do they retire the entire container? Or is it like a memory swap in a standard server; just open the box, replace what's bad and keep moving forward?
Also, what is this going to do to conventional datacenters? If you can functionally build a datacenter in the middle of a field, just by laying a concrete or asphalt slab, and building a perimeter fence...how will that affect datacenters that currently host enterprise level customers? If Facebook were to consolidate facilities into containers and host them in a cornfield in Iowa, what happens to all of the space they occupy on both coasts? Ebay, Salesforce, MySpace, same question.
The datacenter market is very tight, even in the current economy, studies have said the cost of datacenter space will remain where it is (best case), if not go up dramatically (likely) in 2009. With companies halting construction on large scale datacenter builds, thus limiting supply; and companies moving into the cloud, thus increasing demand for large service providers. I can see a quick shift into containers and cornfields.
I don't want to get too far into the cascading effect that could have on the datacenter industry as a whole. It certainly would affect service providers (Silverback included), but it would also affect existing datacenter and co-location companies. If 365 Main, Terremark, Equinix, Savvis, etc lost half of their large customers (30+ racks) to containers. That would free up a whole bunch of co-location floor space, and the all important metric...power.
Again, I am not going to cascade that, but you can start to imagine where that could lead, and depending on your mood, that could be a really good thing...or a really bad thing. I guess what I am saying here is...there are too many unanswered questions with regard to Container Based Computing to truly identify how it is going to impact the datacenter industry as a whole. But it certainly will make an impact, and I can't wait to see how this all plays out!