It's been a pretty information-loaded week regarding energy-management issues within the data center.
The first was the keynote Bill Coleman gave at O'Reilly's Velocity Conference earlier. He addressed why the current trends (point-products, resulting complexity) in data centers is unsustainable, and why economies-of-scale are declining. His answer: the cloud. He talks about 1.0 (where we are today), cloud 2.0 (more sophisticated, may even replace the PC), and 3.0 where it really represents the "webtone" of services available on-demand, reliable, and composable. And the more efficient these economies-of-scale become, the more energy efficient they become as well. DataCenter Knowledge covered that as well, with a few relevant pointers to boot.
The other nifty reference was on Microsoft's TechNet Magazine. There, Dave Ohara pointed out similar observations from a number of sources --- that turning off unused assets is a path to sustainable computing. He gave a few examples including
- Weill Medical College's HPC clusters leveraging IPMI for node shut-down
- Citrix' approach to using their PowerSmart utility to power-down idle HP-based presentation servers
- Microsoft's "Power-Aware Server Provisioning and Load Dispatching for Connection-Intensive Internet Services"
- Commercially-available solutions, such as Cassatt's