More on the Future of Cloud Computing

Posted: 5/10/2008 4:14:00 PM
Geoffrey Fox asked the Cloud Computing Group if Grid Economies would evolve in the cloud computing space. Here are my thoughts on that.
I originally posted this to the Cloud Economies and Standards thread, but I'm adding it here as well. Here's a PDF on Grid Economies with a lot of information about the topic. On with my response...

For standards, I think either a committee will sprout up to manage protocols (better) or the leading supplier will become the de-facto standard (more likely). Boutique providers will distinguish themselves from the pack with support, proximity, price and environmental friendliness among others, which I'll mention later.

As cloud services approach commodity status, they will all have up-time of five nines, about the same performance, network hops, etc, so then price will be a distinguishing factor. Because consumers want a standard pricing metric for commodity products we will likely see grid economies evolving.

Right now pricing is varied and complicated. Amazon's pricing structure is based on server capacity, MediaTemple uses a GPU and Mosso uses requests. As customers already find these pricing schemes too unpredictable, we will start to see tools arrive to help us calculate our application demands in Grid Units. These "Grid Units" could be based on process scheduling priority, processor and network utilization, disk space, memory, etc. "Your application consumes 14 Grid Units." Multiply 14 by the vendor price per Grid Unit and you know what you'll be paying before you move to any particular cloud. That's an oversimplification, but I think the vendors who provide this kind of predictability will win out.

With thousands of customers using the same cloud, eventually demand will outstrip supply and affluent consumers will pay more for quicker apps. I think here, we will see competing vendors catering to different customers by offering different resource management mechanisms. Some customers will want to bid for scheduling priority, others will want to pay a higher fixed rate for higher priority processing. There are lots of algorithms and I don't know which ones customers or vendors will prefer. It is fairly certain though that vendors will not leave revenue on the table when some customers will pay more than others.

Back to standards, I think suppliers using proprietary APIs will present a barrier to service adoption. "Will I have to rewrite my code to work with your storage solution?" If the answer is yes, forget it. When electricity was deregulated, customers could switch their providers to save money or buy greener power, but if they had to rewire their homes or buy new devices -- no one would. ISVs are going to build cloudable products and they'll want to write them once and let the customer choose and pay for the cloud services independently.