Grid and utility operating infrastructures -- now lumped in with "cloud computing" -- can take on many permutations, and are often highly customized using open source and tuned server components. There will be significant and growing use of virtualization at multiple tiers. Hypervisors from XenSource (now Citrix), VMWare, Sun and Microsoft will be common. There are others. See Wikipedia under "Hypervisor" for a nice list and background.
Virtualization should allow support of a large swath of relevant operating system environments in these clouds. Performance has come a long way, and will go further yet.
I also expect that the major software and hardware vendors will be creating packages of "private cloud" infrastructure, which may be more to some enterprise's liking. These will no doubt conform to legacy and commercial deployments, which may make sense for shops that have leaned toward specific for few vendors. They may be able to gain lower TCO without creating a home-growth cloud. These vendors will want to retain your business as private cloud infrastructure providers as they worry about losing you to someone else's cloud -- and perhaps operating environment changes.
For hosting organizations, where every penny and electron counts, they will be highly customizing their secret sauce platforms, and so we won't know too much about what they have under the hood. How much do we know about Google's stack, for example? IBM likes to bring clouds out from their mainframe strengths, Sun from Solaris on UltraSparc, and so on. I expect most of their clouds will be built on Intel/AMD x86 blades, virtualized Linux instances and a variety of virtualized runtime and middleware stacks on top.
Another ingredient that should help differentiate hosts and service providers is the use of MapReduce technology, which is used widely by Google. There is an open source version called Hadoop. These help create a true compute fabric by breaking up compute aspects into nodes and managing those nodes on an amounted and self-sustained basis. See Wikipedia under "MapReduce" for background.
But your question may be moot. The whole point about leveraging grid/utility/cloud services is that you don't have to know or worry about the platform and its ingredients. You should only have to be concerned about the apps and services you require, the service agreements, protecting yourself against risks, and then factoring how much money you'll be saving from hosting your own applications and data on someone else's platforms. Seek the provider that can host your types of applications, provide you assurances on performance and privacy, and then begin to identify those applications you'd rather not produce or host yourself. There will be a mix of hosting options for many years. The right choices on which apps to host how is the key now.
Dig Deeper on Topics Archive
Related Q&A from Dana Gardner
VIrtualization, although not required for SOA, is becoming one of the most important mega trends in the data center right now. Continue Reading
Dana Gardner explains what "private cloud" datacenters are and why they are created by both large and small providers. Continue Reading
Dana Gardner describes three major business strategies for successful business intelligence. Continue Reading