This audio was created using Microsoft Azure Speech Services
The hidden cost of cooling redundancy – and how Data Center Infrastructure Management software helps you identify it.
I recently spent some time in a small data center with a highly redundant cooling setup. Actually, we calculated the cooling redundancy to be N+6 meaning that six of the data center’s in-row coolers could fail without affecting the temperature of the data center.
The reason for this massive cooling redundancy was that the data center was originally designed for more load than was currently deployed – and probably ever will be deployed.
The capital cost of buying the redundant in-row cooling units was of course substantial, but the data center manager wasn’t too concerned about his energy efficiency as the in-rows all have variable speed fans, so their power consumption follow the cooling need. Indeed, the power consumption of the in-row cooling units wasn’t very high.
On examining the data center more closely, however, we discovered that there were actually a hidden – but significant – cost of the oversized cooling system. Although the in-rows didn’t consume much power, they had a subtle impact on efficiency of the UPS.
The UPS in the data center is a modular UPS whose capacity you can adjust by adding or removing power modules. When deciding the capacity of the UPS, the data center manager added up the nameplate values of the equipment in the room, and ensured that the UPS had sufficient capacity to support it. In this calculation he included the nameplate power consumption of all the in-rows in the data center.
By doing so, he oversized the UPS significantly, resulting in the UPS load being just about 20 per cent. At that load, the UPS had a loss of 6.2 percent.
By building a model of the data center in a Data Center Infrastructure Management (DCIM) tool, we were able to determine the maximum load of the in-row cooling units with the current amount of deployed servers, and it turned out we could safely remove one of the 16 kW power modules from the UPS.
Reducing the capacity of the UPS naturally increases the load percentage on the UPS, and with the UPS being more efficient at higher load percentages, the loss dropped from 6.2 percent to 4.8 percent.
This means a reduction of the UPS’s power consumption by 1.4 percent. A number that will quickly make its way to improving the PUE of the data center and reducing the energy bill and carbon emission of the site.
