Vertiv Uses Machine Learning to Automate Data Center Cooling

Vertiv, formerly Emerson Network Power, has introduced a software system that according to the company uses machine learning to automate management of data center cooling systems to improve efficiency.

Data center facilities managers normally have to manage each individual component of the cooling system (i.e. chillers, air handlers, economizers, etc.) to fine-tune the overall system. Change the setting on one, and the entire system gets affected.

The idea with iCOM Autotuning, Vertiv’s new software feature, is to use machine learning techniques to control all of the elements automatically, the company said in a statement.

In a direct-expansion data center cooling system, that means compressors, fans, and condensers are harmonized to eliminate short cycling, which is when cool air returns into the cooling system without going through IT hardware. In chilled-water systems, the autotuning feature avoids rapid fluctuations in valve positions to balance fan speeds, water temperature, and flow rates.

The feature is part of Vertiv’s Liebert iCOM-S thermal system control. It is available for select Liebert cooling systems installed in North America, the company said.

While running in production to improve cloud services by the likes of Google and Facebook, machine learning algorithms are seldom applied to data center management. The rare examples of companies that have done it include Google, which uses machine learning to improve data center infrastructure efficiency; Coolan, a startup that used machine learning to optimize the cost of data center hardware acquired by Salesforce last year; and Romonet, whose software analyzes the cost of customers’ data center assets and traces the impact of infrastructure decisions on their bottom line.