Under HP's new bid-for-compute-time scheme (known as Tycoon) which, for all intents and purposes turns MIPs (millions of instructions per second) into a commodity, perhaps compute power from the company should be listed on Chicago's Mercantile Exchange along with cattle and pork bellies. Or maybe HP CEO Carly Fiorina should drop a line to eBay CEO Meg Whitman to facilitate the bidding. After all, to keep companies like Overstock.com from nosing in on its dominance, eBay is making sure that its getting its fair share of novelty listings (yes, today, bidding for compute power is about as novelty as it gets).
The idea behind this form of utility computing is to use mutli-tenancy (shared infrastructure) in a way that maximizes system utilization (today, a lot of money is wasted on dormant compute power) while properly setting priorities for who gets what compute power when. Not only can savings be derived through the elimination of extraneous overcapacity, the idea of bidding -- especially if multiple compute utility providers are vying for your business -- can drive the cost down. Right now however, with no standards for measuring dissimilar forms of compute power or ways to resolve those dissimilar forms into energy units that can satisfy any need (regardless of what host environment an application is asking for), utility computing offerings are strictly homogeneous and will remain so for quite some time.>
But if the opportunity for a multi-host compute unit does exist, then the closest thing we have to that today, because of the way it works across platforms while insulating the details