Power Usage Effectiveness
data uses more power than disclosed
The Problem With PUE
The PUE ratio was adopted by the hardware manufacturers of storage devices as the best way to show consumption due in part to the limited nature of what is included in the measurement. For Example if you were trying to come up with an energy consumption measurement for a print shop. Would you track the power used only by the printing equipment? Or would you assume that the people working there need lights to see, they need air conditioning to control the heat coming off the big printing machines. They use computers to do the graphics, etc… Don’t all these other things, that are all part of the print shop use power. So if you wanted a real power consumption measurement for the business, then you would need to measure all of the power used by the entire business, not just one specific part, so you could claim to be the most energy efficient print shop. This is the problem with the PUE.
While there have been some technological advancements to storage technology over the last fifteen years. The largest being a much larger usage of flash memory as opposed to spinning disk. All types of data storage drives have gotten bigger and can store much more data while using the same amount of power, so with this increase in data density along with not measuring the total power usage by the entire data center is how operators can claim to use less power today than was used ten years ago.
DC’s continue to increase in size daily. With some Hyperscale DC’s currently being built over 1 million sq. ft. However many are composed of a mix of new and older equipment and lack the innovation required in the storage sector. The industry has solved the perception issue of using “way too much power” by only counting 1/3 of what’s actually used. Current standards are based on flawed calculations and have helped the problem of over consumption explode.
DC’s in the U.S. used more than 160 Billion Kilo Watt Hours (KWh) of electricity in 2019 roughly equivalent to 51 power plants each with a capacity of 500 Mega Watts (MW’s). A typical Data Center in the U.S., which is on average between 10,000 and 30,000 sq. ft., much smaller than the new DC’s in the several hundred thousand sq. ft. range, may consume as much energy in a year as forty thousand U.S. households and use 100x to 200x the energy of a standard three-story office building.
While only using 1/3 of the total DC energy consumption to show the consumers you are being energy efficient is, what we like to call “greenwashing”. Since cooling, servers, other IT, lighting, and HVAC account for 70% of DC’s total power usage. How can they measure energy consumption without including “ALL POWER CONSUMED” by a Data Center?
“There are over 7,500 data centers worldwide, with over 2,600 in the top 20 global cities alone, and data center construction will continue to grow 21% per year into the foreseeable future.”