1 August, 2017
In response to Apple's initiative to power homes with heat generated from data centers....here is our take on how CFD can help design a purpose-built data center for waste heat recovery.
The evolution of the modern data center has been an expeditious and compulsory one. In an extremely short amount of time, the industry has evolved from a chaotic amalgam of mainframes into a multitude of cooling design strategies, operating with a great deal of intelligent design.
More and more power is being crammed into components of ever decreasing size and this trend does not show signs of slowing. This means the efficiency of the heat rejection process is paramount, which in turn, leads design toward more innovative ways of providing cooling. Free air cooling emerges as a sensible option, which more pioneering companies are turning to for future designs.
While many manufacturers can provide units that include air side economizers for a more efficient cooling regime, some are looking at the situation with a wider ambition – utilizing the heat load produced by the IT for waste energy recovery.
A: Exhaust, B: Heat Exchanger, C: Ambient Air, D: Fan Wall, E: Wet Media, F: Return Plenum, G: Exhaust Fan Wall, H: Hot aisle contained whitespace with return exhaust air plenum
The images show an example of one such design. Fan Walls at either side of the facility draw atmospheric air through the servers and back out into the environment, but first-and here’s the kicker- through a set of large heat exchangers. These heat exchangers are the tools by which the energy is retained. They transfer this energy into the water passing through the fins of the HX, which is then pressurized and heated further into steam. The steam drives turbines and the facility manager drives a Ferrari…
The images give an example of how a purpose-built data center can recover the energy lost through IT heat rejection. The structurally intrinsic cooling strategy creates a singular flow path ensuring no mixing of air streams. Without mixing, the exhaust air remains hot and ensures that the maximum amount of energy can be retained by the heat exchangers.
As an example, and with the given ambient conditions, this room can retain 135 kW of heat from the 250 kW dissipated from the IT load. The external air provides the cooling capacity, which means aside from mechanical losses, it is now simply the water pressurization and fan loads that reduce cooling system net power gain.
This proposal, unlike many standard data center designs, has the capability of not only retaining its intended capacity, but increasing it as the business and industry progresses. Many operators find that as their data center matures the potential of it reaching the design capacity diminishes. This is often caused by incorrectly managed IT deployments, due to immediate business driven requirements that overshadow long term cost considerations. Where these less cognizant operators begin to see hot-spots and deem a facility at its limit, their energy retentive, and now perhaps slightly self-satisfied, counterparts can easily add cooling capacity without disrupting their IT supply.
A standard data center design with this problem would require either a fan refit, which can drastically change flow profiles, often in an unintuitive manner or; new cooling units, which in legacy rooms are invariably a costly, yet not very effective solution. Simply adding fans or cooling capacity in the design shown here gives you an identical flow regime, simply with more heat rejection.
The result being that, not only does this type of data center recover waste energy throughout its life cycle; but also, when other operators may need to think about building a new data center, this one can simply trade off the regained energy for more cooling power. Establishing the true cost of a design such as this involves an extensive amount of site specific data, which highlights the need for CFD in the design of such facilities. By using this analysis, an owner/operator can deduce the actual long-term operational costs and make a comparison against more traditional data center designs.
In this example, the waste heat being retained is simply sold back to the local grid; but as an offset to the operational costs, free cooling designs can also be used to heat surrounding offices. This method is more efficient as the retained heat does not need to be converted into electricity before being reduced to a negative value on a costing spreadsheet. This is especially important in the endeavor of retrofitting a system such as this because the relatively inexpensive task of containing hot aisles and ducting hot air into a pre-existing cooling system can result in vastly improved efficiency metrics while increasing actual data center capacity.
CFD is extremely useful in such proposals and can give good insight into the efficacy of a design. It provides a proof of concept under any external conditions, and an understanding of how much the energy retention strategy will yield over the intended life cycle of the data center.
By: Mike Eccles, Consultant Engineer