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.
In our near future virtualization will permit us to disregard where our servers are located in our datacenter (something VMware is trying to approach with VMotion technology).
Then, in a farther tomorrow, virtualization will also permit us to disregard where our servers are located among multiple datacenters, even of different companies.
And so on, up to the final goal of disregarding everything but the application we are trying to use. That application tasks will be processed in a distributed way, among all available datacenters in the Net, which is possibly called Grid Computing (while the pay-per-use of Grid Computing is possibly called Utility Computing).
Benefits of such a distributed environment are huge:
- maximized available resources usage
- increased computation power
- increased resources availability
- maximized high availability
just to name a few.
Virtualization is going in that direction and someone is already trying to reach those results. But there are millions of technical problems to solve, from scheduling to security, from accounting to system management.
And without a common virtualization framework we won’t reach the final goal anytime soon.
So if you asked yourself why VMware is trying to establish virtualization standards this could be the answer: VMware is looking for grid computing much before any other company.
If you feel Grid Computing a fascinating subject you should absolutely read the free IBM Redbook Introduction to Grid Computing.