“Today’s virtualization barely permits us to disregard what kind of resources we really have inside our servers. Companies like VMware, Microsoft, Xen, etc., are offering solutions to share (although still not in a dynamical way) CPU power, memory, storage and partially networking within a single server (or a cluster). Companies like Citrix, Microsoft, Sun, etc., are offering solutions to share applications within a single server (or a cluster). If you think these technologies could be just refined and nothing else, think again.
As network speeds increase, and multicore processors become mainstream (I know it hasn’t happened yet), this seems to be the most logical progression of computing. In grid computing or clusters, the network is just one more bus for information between components. Things like high-end HD graphics require bandwidths that are currently too high to realistically send across a network without a penalty. Other operations can easily be sent across a network trasparently.
I would love to see this available as a workstation/desktop implementation as well as on servers. Imagine if your workstation were merely part of a dynamic partition on a cluster. If the workload increased beyond what your computer could reasonably handle, idle CPU cycles could be used from other computers in the cluster. Imagine a home network where not only files and printers could be shared but idle CPU cycles as well. If burning a DVD, compressing a folder, and downloading some files taxes your computer too much, the spreadsheet you’re working on could be handled by a computer in the next room.