Data Center Power Series – Part 2
Data center cooling and power calculations, while not the most stimulating of conversation topics outside of the IT crowd, is an important and often discussed component of data center design, efficiency, and scalability. The purpose of this whitepaper is to provide a clear and concise guide to the data points and calculations needed for determining and understanding the relationship of voltage, amps, watts, and BTUs. With a few basic formulas and an understanding of the data needed to perform the calculations, you can ensure an efficient and functional data center implementation.
Start With the Correct Data
Before performing any calculations for data center cooling, it is critical to determine the correct input data. For the purposes of this document, let’s assume we are adding equipment to an existing data center environment and need to calculate the power and cooling requirements to determine if our data center has sufficient cooling capacity, typically expressed as Tons of Cooling.
To calculate the cooling requirements we first need to lookup the environmental specifications (not the power supply wattage rating) for the devices to be installed and determine the expected operating amps for the devices at the voltage being supplied to them. Using voltage and amps, we calculate the output watts as a measure of heat energy to calculate cooling requirements using the formula Volts x Amps = Watts.
In our data center scenario, we’ve determined that the highest voltage our servers and network switches are both capable of running at is 240 Volts A/C. Since higher voltage is more efficient in terms of data center power delivery, we will supply 240 Volts A/C to our new racks. (For more on power selection and efficiency, including the advantages of 240V over 208V, see Part 1 of this series titled, Data Center Power and Cooling Series – Part 1 – Choosing the Right Voltage.)
Examining the spec sheets, we find that our servers operate at 2.375 Amps and our network switches at 0.700 Amps when running at 240 Volts A/C. In our scenario, we will be adding 20 4-post racks in 4 rows of 5 racks each in a hot aisle/cold aisle configuration (see figure below) and will populate the racks with 400 2U servers and 40 1U 24-port network switches. Equipment is densely distributed with 20 servers per rack and 2 redundant network switches (rear mounted) per rack.
Using standard conversion equations (Watts x 3.41 = BTU/Hr) and (BTU/Hr / 12,000 = Tons of Cooling) we can calculate the amount of data center cooling required. (Interesting Fact: In terms of cooling, a Ton correlates to the amount of heat energy required to melt one ton of ice at 32 degrees Fahrenheit in one hour.)
Now we can calculate total Amps required per Rack, and total Watts for the cooling calculation.
** In an actual data center scenario, we would determine the total Amps required per rack and then employ a calculation to prevent over drawing Amps on the circuit during power spikes such as at server startup times. Using the 80% rule for data center power utilization and assuming staged startup cycles for the servers to avoid excessive amperage draw, 60 Amps per rack should be sufficient in the above example ⇒ 60A x 80% = 48A > 47.5. In many cases, data centers use the maximum available Amps value (60A in this case) to calculate cooling requirements and thereby provide a buffer zone for cooling load spikes as well.
Now we have the data needed to provision appropriate circuits to the racks to power the planned equipment expansion project and the cooling requirements for the new equipment. The next step is to evaluation current data center cooling to determine if the new equipment will fit within the existing cooling capacity. If the current data center cooling capacity falls short of our needs in the planned deployment area, we will need to provision additional cooling prior to implementation.