We’ve covered previously in this blog the different data center cooling architectures – room, rack and row-based cooling. But we recently came out with an updated white paper that helps customers determine which architecture – or mix of architectures – makes sense for them.
I co-wrote the white paper (no. 130, “Choosing Between Room, Row, and Rack-based Cooling for Data Centers”) with Kevin Dunlap. In it we explain how it’s essential to relate the performance characteristics of the cooling methods to practical issues that affect the design and operation of real data centers.
We compared the three cooling methods against 11 criteria that are important to data center owners and operators, namely:
- System availability
- Life cycle cost (TCO)
- First cost
- Electrical efficiency
- Water piping or other piping near IT equipment
- Cooling unit location
- Heat removal method
For the first five criteria, we lay out the pros and cons of rack, row and room-based cooling options. To summarize, we found that rack-based cooling is the most flexible, fastest to implement, and achieves extreme density, but costs more. Row-based cooling provides many of the same advantages of the rack-based approach, but with less cost. Room-based cooling allows for quick changes to the cooling distribution pattern by reconfiguring the floor tiles and allows cooling redundancy to be shared across all racks in the data center with low densities. This method offers cost and simplicity advantages.
We looked at the remaining six criteria one by one and came to some interesting conclusions. With respect to first cost, for example, we found that room-based cooling was least expensive, as you’d expect. But as densities grow, the other options become less and less expensive, such that savings in other areas may eventually offset a higher first cost.
One of those other areas is electrical efficiency. Electrical costs for room-based cooling without using hot-aisle containment were highest, but decrease slightly as density increases. Row-based cooling helps reduce unnecessary airflow, which can save more than 50% of your fan power consumption as compared to room-based cooling. Costs naturally go up as rack density increases although adding redundant units will actually lower the energy consumption but will result in higher first cost. With rack-based cooling, energy costs are higher at lower densities but go down as densities increase – at least to a point. At high densities (above 12 kW per rack), energy costs start to climb again because each rack has its own cooling unit and fans approach their maximum operating speed to provide the necessary airflow, among other issues.
We also found the location of an air conditioning unit can have a dramatic effect on system performance, although this is an issue mainly with room-based cooling in data centers with no containment systems. For row-based systems, we have design rules that tell us the best locations for air conditioners. And for rack-based systems the issue is eliminated because you can determine the exact location of the air conditioner with respect to the target load, although future additions require some planning and forethought.
As the paper makes clear, there’s a place for room, row and rack-based cooling systems. The trick is in knowing which is best for your data center, both now and in the future, and to plan accordingly. Check out Schneider Electric white paper no. 130, “Choosing Between Room, Row, and Rack-based Cooling for Data Centers,” to learn more.