Back in 2005 Sun Microsystems announced the company planned to build the world's first online compute exchange. More then three years later there has been no other mention of this supposed compute exchange. In the original press released they described the offering as a plan to introduce a new electronic trading environment that will allow customers to bid on CPU usage cycles. They went on to say that being able to dynamically bid for open compute cycles will provide companies across the globe with unprecedented flexibility in planning for the purchase and use of compute power. This is a new paradigm in computing where companies can access an unlimited number of CPUs as they need them. Today as cloud computing begins to take off and regional cloud utilities start to come online the idea of a cloud exchange is again beginning to be discussed.
Back in April at the Interop conference several attendees mentioned they wanted to see the creation of such an exchange platform. The reasoning was that as new regional clouds come online having a uniform point of entry to a world wide cloud ecosystem will make this type of transition more efficient.
Right now most clouds have there own set of APIs, interfaces and unique technologies. An open compute exchange may provide a centralized point where cloud consumers and providers would be able to make decisions based upon which cloud resources they may want to utilize as well as a clearing house for providers with excess capacity. Variables may include metering based on actual use of the resources in CPU hours, gigabits (Gbs) consumed, load, network I/o, peak vs off peak time frames, geographical location, SLA's, and quality of service rules could be just some of the metrics that determine the price of a cloud providers resources.
One usage example might be in terms of a green or eco-centric point of view. Let's say Cloud A uses cheaper coal based energy source and Cloud B uses a more expensive Hydro source. Although more expensive, choosing Cloud B may help offset an enterprises carbon credits and therefor actually be a bit cheaper from a carbon point of view.
Another example may be based on geographical cloud computing. Let's say a UK cloud and a North American cloud. Rather scaling based on system load, a cloud user may want to monitor application response times based on geographical location and scale according to an end users experience. By have the option to access compute capacity through an exchange, cloud consumers who are running global network services would no longer have to signup for several cloud services. This would also effectively render edge based CDN services like Akamai irrelevant.
What are your thoughts?