Realtime Bidding for Cloud Computing

I recently read an article in the Economist and then a few other blog posts on an emerging concept of offering cloud computing processing and content delivery on a bidded pricing scheme. The basic idea is what is being called a “follow the moon” strategy. This is based on the fact that in geographic areas where it is night time or a cooler environment, computers and servers are either sitting idle or need less energy to cool. You may have heard that Google is setting up data centers on mountains or in the arctic so that they don’t have to pay for air conditioners to cool down their servers and can do more data processing and storage for cheaper.

Well with the dawn of virtual machines and a software layer that splits the application and processing algorithms from the actual hardware that does the storage and processing it is now possible to float your application and load balancing from cloud to cloud to get the best price.

Amazon is already offering a service they call “Spot Instances” this reminds me a lot of RTB (Realtime Bidding) in the ad world. People are starting to choose to run a server instance and have more content delivery power based on the price to do so. For someone that is streaming a ton of video and costs could expand rapidly can technically build this into their revenue model and essentially choose to serve customers based on the current going rate to deliver that content or compute that request.

Amazon offers bidding through an API and with their DevPay option I think it would be interesting to offer a service layer to optimize pricing over time and estimate yield management from an ad supported site. For example, lets say you use Rubicon Project for your SSP to control yield management on your site and maybe you’re promoting your site out there with ads on a DSP and you’re using an SEM to bid into search. It would be interesting to build a piece of middle-ware between these to estimate yield management on what you think you are going to make on a server call and process based on how much you can make from your SSP and the n+1 of acquiring an additional user via your DSP and SEM and essentially decision on that in realtime…..or near realtime….or even hourly for that matter.

Currently AppNexus is offering a seemingly open source stack to do this but adding in the yield management on the computing costs could essentially put you in the driver’s seat of controlling your cost model, revenue model, and better estimate your growth plans. Basically you’re just left with the Marketing Idea, and the Business Model itself!

Here’s an interesting site that tracks the price of Spot Instances on Amazon.

http://loca.ly/dcBwAg