A leading data center engineer at Google Inc. (NASDAQ:GOOG) – Jim Gao, nicknamed ‘Boy Wonder’ – has developed an amazingly innovative solution for monitoring and handling the company’s power usage efficiency levels (PUE) across its data center network. Following his participation in an online class on machine learning, Gao developed a model by which PUE data, which incorporates a mass of information operative across 19 different variables (cooling tower speed, outside temperature, and so forth), could be systematically monitored and scrutinized.
The importance of Gao’s model cannot be understated. Firstly, the machine learning system is capable of instantly alerting Google’s data center engineers when data falls outside of its prediction scope and thus indicate when something is wrong with a center, allowing them to respond accordingly. It also allows Google to oversee its power usage in a far more efficient and well-thought out way. For instance, the data produced can be used to identify the most appropriate and cost-effective time to clean a data center’s heat exchangers, saving both invaluable time and money.
Previously, data center operators were witnessing approximately half of their energy usage taken up by cooling equipment, but of late, leading companies such as Google and Facebook have implemented measures such as the use of external air for cooling, or operating their server facilities at warmer temperatures, in order to limit energy use.
Before Gao’s invention, all the analysis required to direct and inform such schemes had to be carried out by humans looking at an ever proliferating mass of spreadsheets – a mission impossible scenario in terms of developing appropriate strategies and responses for dealing with PUE levels.
