Decision Support Questions


1. What is the average kWh change per degree Fahrenheit for the most extreme temperature differences in the study period?


For each increase in degree Fahrenheit the expected change would be an increase of 39.75 kWh.

Read More

For the calculation of this number, please see the following:

Temperature Graph
11-22-13 11-24-13 Temp. Difference
Avg Temp. 54.66 26.08 28.58
Avg Low 51.00 23.30 27.70
Avg High 58.10 32.50 25.60
11-22-13 11-24-13 Energy Difference
Total Power 6375.69 5242.35 1133.34
IT Power 2841.58 2850.38 -8.80
PUE 2.24 1.84 0.40
\[ average\ kWh\ per\ degree\ Fahrenheit = \frac{Energy\ Difference}{Temperature\ Difference} \]

Where energy difference is Total Power on 11-22-13 minus Total Power on 11-24-13, and Temperature Difference is Avg Temp. on 11-22-13 minus Avg Temp. on 11-24-13.

\[ average\ kWh\ per\ degree\ Fahrenheit = \frac{6375.69-5242.35}{54.66-26.08} \] \[ average\ kWh\ per\ degree\ Fahrenheit = \frac{1133.34}{28.58} \] \[ average\ kWh\ per\ degree\ Fahrenheit = 39.65 \]

2. According to the results in question 1, what would be the effect of a 1 degree Fahrenheit increase due to climate change on this case study data center?

In this situation, the data center would need to use 39.75 kWh of power to keep the internal data center temperature at the same level.

3. What is the average kWh per degree Fahrenheit for two days with similar or trending outside air temperature?


For each decrease in degree Fahrenheit the expected change would be an increase of 462.80 kWh. This extreme difference between question 1 demonstrates that there is still a need to obtain a baseline for the fan system that uses the greatest share of energy usage.

Read More

For the calculation of this number, please see the following:

Temperature Graph
11-14-13 11-21-13 Temp. Difference
Avg Temp. 40.99 46.72 5.73
Avg Low 32.8 37 4.20
Avg High 48.7 54 5.30
11-22-13 11-24-13 Energy Difference
Total Power 8102.35 5450.45 2651.90
IT Power 2822.75 2789.10 33.65
PUE 2.87 1.95 0.92
\[ average\ kWh\ per\ degree\ Fahrenheit = \frac{Energy\ Difference}{Temperature\ Difference} \]

Where energy difference is Total Power on 11-14-13 minus Total Power on 11-21-13, and Temperature Difference is Avg Temp. on 11-14-13 minus Avg Temp. on 11-21-13.

\[ average\ kWh\ per\ degree\ Fahrenheit = \frac{2651.90}{5.73} \] \[ average\ kWh\ per\ degree\ Fahrenheit = 462.80 \]

4. According to the results in question 3, what would be the effect of a 1 degree Fahrenheit increase due to climate change on this case study data center?

In this situation, the opposite occurs in that the data center would use 462.80 kWh of power for each degree cooler. While this situation is not logical, it demonstrates the variance in the data.

5. If we downsize our data center and reduce by one half the number of servers by shifting traffic to a cloud provider. What is the effect on the carbon footprint?


Therefore, the carbon footprint on a typical day would be reduced from 3.88 metric tons per day to 2.28 metric tons per day.

Read More

For the calculation of this number, please see the following:

Assuming that the cooling would be also cut in half and that the fan and pump would stay the same, the result is the following:

\[ CF = Total\ Energy\ Usage\ in\ kWh * 1.34 lbs/kWh*1 metric\ ton/2204.6lbs \]

Where CF equals the Carbon Footprint, Total Energy (TE) equals the energy used by day for the cooling system and the PDUs, 1.34 lbs/kWh equals the national average for carbon emissions, and 1 metric ton equals 2,204.6 lbs

11-21-13 Before After
Total Energy (TE) in kWh 6375.69 3756.99
TE*1.34 8543.43 5034.37
Carbon footprint in metric tons per day 3.88 2.28

6. From the data in question 5, what would be the percentage decrease in the carbon footprint?


As presented above the percentage decrease of the carbon footprint per day after the above stated assumptions is that there would be approximately a 41% reduction. While this figure is specifically from one day as a sample, other days were found to be in the same range. Also this figure does not represent the shift in the burden, or in other words, the carbon footprint of the servers that are now hosted on the cloud.

Read More

For the calculation of this number, please see the following:

\[ \% \Delta CF = \left( \frac{CF\ Before - CF\ After}{CF\ Before} \right) * 100 \] \[ \% \Delta CF = \left( \frac{3.88-2.28}{3.88} \right) * 100 \] \[ \% \Delta CF = 41 \% \]

7. Using question 5 again, what would be the effect on the PUE when the number of servers are reduced by half?


An interesting finding is that there appears to be a general trend where the PUE actually increases when the number of servers is cut in half. The assumptions would need to be further tested, specifically what would the fan and pump energy usage be after the decrease in the number of servers?

Read More

For the calculation of this number, please see the following:

\[ PUE = \frac{Total Facility Energy}{IT Equipment Energy} \]

Where Total Facility Energy equals the summation of PDU, AC, Pump and Fan energy usage, and IT Equipment Energy equals PDU energy usage.

11-21-13 Before After
Total energy 6375.69 3756.99
IT Energy 2841.58 1420.79
PUE 2.24 2.64

8. What would be the percentage increase in the PUE after the number of servers is cut in half?


There is a general trend of the PUE increasing by approximately 18% when the number of servers is cut in half. This figure as stated above would be adjusted when the fan and pump energy can be better predicted.

Read More

For the calculation of this number, please see the following:

\[ PUE = \frac{Total Facility Energy}{IT Equipment Energy} \]

Where Total Facility Energy equals the summation of PDU, AC, Pump and Fan energy usage, and IT Equipment Energy equals PDU energy usage

\[ \% \Delta CF = \left( \frac{PUE\ After - PUE\ Before}{PUE\ Before} \right) * 100 \] \[ \% \Delta CF = \left( \frac{2.64 - 2.24}{2.24} \right) * 100 \] \[ \% \Delta CF = 17.85 \% \]

9. What would be the percentage change in energy consumption by lowering the internal temperature of the data center by 2 degrees Fahrenheit?


In this example while outside temperatures where approximately equal, it was found that by lowering the temperature in the data center resulted in a slight decrease in energy usage of 35 kWh. The PUE remained the same in both cases.

Read More

For the calculation of this number, please see the following:

12-2-13 (Initial) 12-4-13 (Lowered) Change
Total Power in kWh 5706.60 5671.58 35.02 kWh less power used
IT Power in kWh 2852.23 2841.10 11.12 kWh less power used
PUE 2.00 2.00

10. What is the typical fluctuation in the PUE over the study period?

We have found that after adjusting the energy usage of the pumps and fans that are one of the primary sources of electrical consumption variance, that PUE in our data center is around 2.0. For a better visual representation of the data please see our website here.

11. What is the maximum server consolidation possible?



Read More

Please see the Systems for the maximum system utilization. The maximum CPU-busy and Memory-busy values were used as input for an algorithm that aimed to find a reduced number of systems by consolidation. For example, if initially:

  • host1 has a maximum CPU utilization of 40% and a maximum memory utilization of 30%
  • host2 has a maximum memory utilization of 50% and a memory utilization of 20%
  • the two hosts could be consolidated in one with potential processor utilization of 90% and maximum memory utilization of 50%. The approach assumes similar properties among the systems. Additional parameters need to be employed if the systems are not equivalent.

    Using data available for this server room, the following consolidation is possible: