Passive vs Active Rear Door Cooling- Which way and why? Is there a compromise?
It's the age old question in the pursuit of the best possible PUE, but the answers may not be as simple as it first appears. Can the new HPC environment provide some hope moving forward?
On paper, removing the need for power and, achieving the same level of cooling feels like a win win option. You get the same functionality from the active equipment in operation and cut a significant portion of power used for cooling, all whilst maintaining the cooling required for the space. Its the perfect area where the PUE can edge ever closer to that illusive number 1.0
Issues.
The difficulty however can be far greater than the theory suggests. The main problem with Passive systems has always been the capacity of the fans to deliver pressurized air. Lets explore that a typically accepted static pressure provided by fans within active equipment is not the best, even at the point of generation. Each component this air now passes causes a drop in pressure. This is until the air reaches an object which requires more static pressure to pass than it currently possesses. At this point the air will stop flowing and gather.
In a passive rear door system this area is typically the coil (the coil is where the heat transfer occurs and allows the air, after the heat is absorbed, to pass back into the room at ambient temperature) and as a result the system no longer functions as designed or intended.
A second problem traditionally is the issue of monitoring. The first time some people will have knowledge of an issue is when the active equipment, is at the point where, it is already heading towards its maximum operating temperature. Therefore this gives a very limited window to be able to respond before any damage occurs.
Both of these problems are big but the third can make a real difference to the overall efficiency of the room in question and that is control. Control follows monitoring because theoretically if you can monitor something then you can also learn to control it. In a data environment the control on a typical door can be anything from the speed of the fans through to the volume and velocity of water being circulated within the coil and subsequently the efficiency of the pumps on site.
Is HPC the answer for passive?
One of the most convenient byproducts of HPC is the requirement for the active equipment to be able to move more air and therefore dissipate additional heat which it has created. This has created a system where given the right questions on design, placement and requirement a passive door can function efficiently and correctly.
Questions;
These can be as simple as how densely populated is the rack in question? A sparsely dense or incorrectly blanked rack may end up recirculating air as opposed to pushing it through the coil.
Placement and spacing could be spoken about before hand as, if the option is there, then the equipment being placed as close to the coil as possible with little interference is the best way to solve this problem. Knowing the static pressure and heat load from the active equipment is the best possible situation.
As a manufacturer we know the pressure drop across our coils and the volume of air required for a specific kW rating. If your requirement is for a 40kW load which would generate an air pressure drop of 55pa across the coil but the active equipment is only capable of delivering 40pa of pressure then this solution is not viable from the start and an active RDHx should be explored instead.
How can you alleviate these issues?
If from the outset you aren't sure of the pressure available from the active equipment and you are working in an environment where high heat loads are probable, but without the ability to experiment after installation, then an active door could be the best way forward as it provides the safety of knowing the fans are capable of air exhaust.
Why active over passive?
In all the scenarios above the active door would allow for prevention rather than cure. A fully functioning active door with the correct water flow and temperature is capable of performance in excess of 0-100kW load.
Another benefit is that with the door being the primary air extraction method is that it allows for the active equipment to run more energy efficiently. This is because heat is removed quicker and therefore fans will not increase in power usage as the temperature increases.
The main caveat to this with the active solution is the volume of air generated by the active equipment. If this is greater than the door can generate then a fan stall can occur, this could be on the active equipment or the rear door itself.
Is there a middle ground?
Fortunately at USystems we understand that not all data centers are the same and do not all conform to an out of the box approach. This is why we have developed our new product the 'ColdLogik CL20 Smart Passive' a rear door heat exchanger but with a modular upgrade philosophy.
This allows, with the correct design, a passive door to be selected for initial implementation and then upgraded in a total of 4 steps unit it becomes a fully active door in the same vein as the rest of our CL20 range.
The main benefits to this are the ability to ,first of all, install a passive door to try and utilize better energy efficiency at lower density, when this passes to a point when pressure drop is too large or heat dissipation to great, you can upgrade to have a definitive kW load of which the door will be capable of (based on individual site conditions of which calculations are evaluated from).
The other advantage to this is that it allows you to upgrade with definitive upgrade steps over a period of time or as we like to say 'Pay as you grow'. In addition to allowing your rear door to grow with your data center capacity, you also avoid the painful experience and cost of ripping out existing solutions to install new cooling equipment.
Taking this route also ensures no waste is created by removing the old system and transporting both the new and old to and from the site.
The CL20 Smart Passive will be launched in Q1 2020 however if you have any questions then please do not hesitate to get in touch. Make sure you come speak to us at SC19 in Denver where the Smart Passive will be on display.