News

Microsoft Working To Squeeze More From the Cloud

• By Kurt Mackie
• March 02, 2009

From that premise, Microsoft aims to improve datacenter efficiencies for its Windows Azure cloud computing and online services offerings by optimizing the datacenter's hardware and software.

One of the group's ideas is to reduce power consumption in the datacenter by essentially clustering a bunch of laptop computers. CCF is experimenting with using Intel's Atom processor, used in netbooks, because of its huge power savings.

"CCF has built two server clusters using low-power, Intel Atom chips and is conducting a series of experiments to see how well they support cloud services and how much their use can reduce the power consumed by those services," Reed explained in the interview.

Atom can use 5 watts of power vs. 50 to 100 watts used by other processors typically found in the datacenter. The Atom processor's power savings from its sleep mode is also a benefit.

To utilize that sleep mode, the CCF developed a solution called "Marlowe" that calculates when to put the computers to sleep. The task is made tricky because "it takes 5 to 15 seconds to awaken a processor from sleep and 30 to 45 seconds for hibernate," Reed explained.

He added that "the system needs to hold some processors in reserve and to anticipate the workload 5 to 45 seconds in the future to ensure that sufficient servers are available."

CCF is also working to create software specifically designed for the datacenter, such as Monsoon, a more simplified communications protocol. A paper (PDF) on the topic describes Monsoon as "a blueprint for commoditizing the networks of data centers used for 'cloud' services where large numbers of servers cooperatively handle huge workloads."

Reed also described a software platform called Orleans that is designed to make building services easier. The Orleans platform runs on Microsoft's Windows Azure operating system in the cloud and "provides the abstractions, programming languages, and tools" to build the services.

Datacenters were originally built using off-the-shelf technology, but it wasn't very power efficient, according to Reed.

"One common analogy is that if one built utility power plants as we build data centers, we would start by going to Home Depot and buying millions of gasoline-powered generators," he explained.

That older approach to the datacenter has run its course. By its research, the CCF team hopes to reduce power consumption four-fold or greater, according to Reed. The group also expects to see lower cost commoditized datacenter solutions as a result of its investigations.

For more information, see the CCF home page here.