The news comes at a time when the cloud computing concept is gaining traction. Late last year Microsoft released a beta version of its cloud service called Azure, following similar services pioneered by Amazon.com and Google.

Earlier this month, researchers at the University of California at Berkeley released a white paper on cloud computing, calling for a standard applications programming interface to drive growth in the emerging sector.

"I believe we're far too early in the process to encourage or enforce standardization," said Dan Reed, strategist for scalable and multicore computing at Microsoft Research who oversees the new R&D team. "We need the corporate and academic communities to continue to innovate in this space because it is evolving rapidly and I expect new types of applications to continue to emerge, combining new and existing services and capabilities in unexpected ways," he added.

The R&D effort has been proceeding quietly since before Reed joined Microsoft in September 2007 and is currently hiring additional researchers. "Our charge is to take a blank-sheet-of-paper look at hardware and software needed to support cloud computing services to see if we can achieve efficiencies with new design points," Reed said.

For example, tomorrow's applications will be able to shift from running on client systems to servers to data center clouds, he said "You'll see a whole new generation of apps emerge that don't look much like existing clouds," he added.

The team will host demos of two of its projects at the TechFest event.

In one demo, researchers will show a server based on Intel Atom chips can handle some workloads at lower power and higher utilization rates that traditional X86 server CPUs. Another demo will show a way to automate server resource management through real-time performance measures and machine learning.

Separately, the group is studying issues such as the impacts on data center computing of emerging technologies such as flash memory storage and on-chip optical interconnects. It is also studying how to optimally use geographically dispersed data centers for cloud services.

"The interesting thing is looking at the interactions of all these things," said Reed. "We have a whole bunch of knobs we can turn and need to find the right applications for them, so we need to build prototypes of hardware and software to test the ideas," he added.