One of Microsoft's labs created a video on measuring power consumption on desktop devices. This video is appropriate for someone who is new to measuring energy consumption on the desktop.

So how much power do your PCs draw?  And how do you figure that out?
Those were the questions that I asked Grant after he recently updated the machines in some of the classrooms he manages.  His problem was even more complex:  He needed stronger/better/faster machines, but he was at capacity on his circuits, so he needed to do it without increasing power draw.
In this video, Grant walks us through some of his methodology in measuring and benchmarking power consumption on a few machines, and shares the results with us. 
It leads to some interesting conclusions, and some good food for thought for anyone trying to make energy conscious decisions around PC purchases.

In this video they use the Watt's up device, but hopefully, they'll put the Smart Watt device in their small server room.

Microsoft Research has another paper on creating a Green Data Center by turning off servers. The paper has been well received and the authors will present at USENIX Networked Systems Design & Implementation conference in April 2008.

The specific scenario they have studied is Microsoft's Windows Live Messenger Servers, combining the challenges of a connection-intensive Internet service (Windows Live Messenger) with turning off servers.

Energy consumption for hosting Internet services is becoming a pressing issue as these services scale up. Dynamic server provisioning techniques are effective in turning off unnecessary servers to save energy. Such techniques, mostly studied for request-response services, face unique challenges in the context of connection servers that can host a large number of long-lived TCP connections. Such servers usually limit on how many new connections they can accept per second and, therefore, a server cannot be fully utilized immediately after it is turned on. Moreover, before a server is turned off, all its active connections need to be reconnected or migrated to other currently active servers. In this paper, we characterize unique properties, performance, and power models of connection servers, based on a real data trace collected from the deployed Windows Live Messenger. Using the models, we design server provisioning and load dispatching algorithms and study subtle interactions between them. We show that our algorithms can save a significant amount of energy without sacrificing user experiences.

The simulated energy savings ranged from 20 to 30% depending on the server-initiated disconnection rate.

Microsoft Research also published a paper on saving Disk Energy in a data center.