The Spike that Kills
Nick Carr notes that the kinds of periodic spikes in traffic that companies sustain (Intuit at Tax time, retailers near Christmas, flower vendors near Valentine's day, etc) result in an over-buy of capacity for most of the year's traffic load:
To run its business with private, dedicated servers, Intuit needs to build its data centers with the capacity necessary to handle the extreme spike in traffic - the peak load - that comes on tax-filing day. Thge vast majority of that installed capacity will go unused most of the time. Multiply that low capacity-utilization rate across thousands of companies, and you get a good sense of the wastefulness inherent in the proprietary model of computing, particularly as companies have to handle rapidly fluctuating web traffic. The only way to do cloud computing efficiently is to share the cloud - to establish a broad, multitenant grid (or a number of them) that balances the loads of many different companies. Otherwise, it'll be one cloudburst after another, and a whole lot of underutilized capital assets.
I wonder if Amazon has sales staff over at Intuit right now, pitching EC2? It looks to me like Amazon may have been crazy like a Fox with that initiative - because it's exactly the kind of "utility" grid that Carr is on about.
