This is the second and final part of the Back to Basics: Cooling – Part 1 article.
Here we will cover:
- Relative Humidity
- Calculating Cooling Capacity
- Converting Between Units
For handy references and info, download my Power & Cooling Cheat Sheet to the right
Approx Reading Time: 10-15 Minutes
The datacenter may seem like a bad place for water, but without its presence in the air, some bad things can happen. An atmosphere with a low relative humidity can cause the air moving through a server to build up static charges on the circuit boards, which can cause damage to electronic components when discharged. On the flip-side, an atmosphere with a high relative humidity can reach the dew point and cause water condensation on equipment creating rust and electrical shorts.
Relative humidity (RH) is the ratio of water mass in the air over the amount of water the air can hold at a particular temperature. As the temperature of the air increases and decreases, the relative humidity decreases and increases respectively (it is inversely proportional), assuming the mass of water vapor in the air stays constant. RH is measured as a percentage where 0% is absolutely no water in the air and 100% is saturated air at the dew point (for its temperature).
This flux of relative humidity is why you will typically see dew on grass and plants in the morning. The amount of water in the air increased throughout the day as the temperature rose. When temperature decreased during the night, the amount of water in the air stayed the same as the temperature fell; causing the RH to increase. When the RH hit 100%, it reached the dew point and the water in the air condensed onto objects surrounded by the air (notably: grass and plants).
Generally speaking, the recommended datacenter atmosphere is between 65°F (18°C) and 80°F (27°C) with a relative humidity between 40% and 60%. This recommendation may become defunct in the future due to an evolution in the industry to move to IT equipment which can withstand higher temperatures; allowing datacenters to cool themselves using fresh outside air instead of air conditioned by expensive and energy hungry systems.
Calculating Cooling Capacity
This is where the rubber meets the road. Capacity planning is part mathematical calculations: adding up all power which will be consumed then calculating what kind of cooling will be needed to remove the heat waste, and is also part art: guessing what the facility will require in the future as technology and requirements change. Since the “art” part of planning can get a little wonky, let’s start with the math.
The best place to start when calculating required cooling is power draw. With very few exceptions (like POE) all power consumed by IT equipment is turned into heat. Since the electrical signals coming into and leaving the datacenter are negligible, and IT equipment doesn’t output any kind of mechanical work, it is safe to assume that all power is being turned into waste heat. This means if you know how much power each device will consume, then you know how much heat it will produce. Face it, your datacenter is nothing but a big, expensive heater.
Figuring out your total wattage (power) draw can typically be done by looking at the capacity of your UPS system. If your UPS can only output 10,000 Watts/VA (there is a difference, but suffice it to say: it is negligible for IT equipment), then you can start at 10,000 watts (10kW) for your cooling calculation. We will use this as a starting point in the example calculation.
UPS Max Output = 10kW
Expected UPS utilization: 50%, or 5kW of draw by equipment
After determining we only expect to draw 5kW of power, we can add in some additional metrics:
(1) UPS Inefficiencies: (.04 × UPS Max Watt Rating) + (.05 × expected wattage draw)
(2) Power Distribution: (.01 × UPS Max Watt Rating) + (.02 × expected wattage draw)
(3) Lighting: (22 Watts × Floor Area in m²) or (2 Watts × Floor Area in ft²)
(4) People: (100 Watts × max number of people)
Saving you the long form math, considering 100ft² of space and no people, we end up with 6050 Watts of heat production.
Now we get into the “art” and the more elusive metrics. It is obvious that much of the heat gained or lost by a facility is due to the environment around it. Depending on the insulation in the external walls and the climate, this factor can be orders of magnitude higher or lower. Since this factor is, by itself, very complex and tricky, we will leave it out of the discussion for now. It is also important to note that if your datacenter is an air conditioned room inside of an already climate controlled building (like a data room in an office building), then you can, for the most part, ignore this metric.
The other evasive metric is utilization change or “room for growth”. We scoped a 10kW UPS but only expect to use 50% of that max capacity. If we scope a cooling system which perfectly matches the expected 6050 watts of heat production, then we have no room for growth in regards to cooling. We can scope a cooling system which matches the UPS, but that can get expensive. This metric varies widely depending on the particular installation. My advice is to calculate the cooling needed for initial expected consumption, then calculate the cooling for the max capacity of the power systems, and pick a size in the middle based on your best guess of required headroom.
Now we have determined we need to account for 8000 Watts of heat production (6050 + some growing room), we can convert this to a number which is meaningful to a HVAC engineer. Using the conversion tables above, we can determine we need 26296 BTUs or 2.272 tons of refrigeration capacity. Since refrigeration units come in integer sizes of tons, or factors of 12,000BTU, we have to round up to a 3 ton or 36000 BTU unit. And there you have it.
In the end, remember you are [probably] not a mechanical or HVAC engineer. The information in this article is not intended to train you to do these calculations and completely scope an HVAC system by yourself, but you now have a pretty good idea about how an expert will go about scoping the system and getting a power consumption estimate from you will probably be part of his process. Knowing details of the calculation process will enable you to better collaborate with the architect of the environment control systems.