Google Turns to Machine Learning to Build a Better Data Centre

me@ryanorban.com (Ryan Orban) — Wed, 28 May 2014 00:00:00 +0000

Summary

This 2014 ZDNet report covers Google’s early use of machine learning to optimize data center operations — specifically cooling system efficiency. The disclosure was notable at the time because it was one of the first public acknowledgments that Google was using neural networks to automate infrastructure decisions that had previously required expert human operators.

Data Centers Waste Vast Amounts of Energy

me@ryanorban.com (Ryan Orban) — Sun, 23 Sep 2012 00:00:00 +0000

Summary

The NYT’s James Glanz reported in September 2012 that data centers were consuming enormous amounts of electricity while using only a fraction of their capacity. The industry figure that stood out: servers typically ran at 6–12% CPU utilization, burning full power while doing almost nothing. The cause was a structural incentive problem — IT departments were penalized far more for downtime than for energy waste, so they grossly over-provisioned capacity and left machines idling rather than risk being under-resourced.

Data-Centers on Ryan Orban

Google Turns to Machine Learning to Build a Better Data Centre

Summary

Data Centers Waste Vast Amounts of Energy

Summary