Since a data center provides information (e.g., data) from a source to a sink in its operation, the waste factor can be applied to gain new and vital insights to optimize the data center power efficiency. W can be applied to a data center if certain assumptions are made about how and where the signal power is transferred. Assuming that Equation 39 defines PIT as the power consumed by the IT equipment, (e.g., signal path power and wasted power on the signal path) and Paux is the auxiliary equipment power consumption, then Equation 39 may be recast in terms of W. This use of W considers only the powers consumed and delivered from a source to a sink within a data center while treating Paux as off-path power that is not involved in the computation of W per Equation 9. Total data center power consumption is then computed through the sum of signal-path and non-path powers.

In the data transmission or processing within a data center, the major power consumption is attributed to servers, network switches and computing equipment, while additional power consumption is associated with cooling systems, power distribution units (PDUs) and other auxiliary equipment. According to findings,9 servers and networking equipment account for about 60 to 70 percent of the overall power consumption in a data center. Cooling systems contribute about 30 to 40 percent of the total power consumption, while the remainder is consumed by power distribution units (PDUs) and other auxiliary equipment.

To break down the information path power consumption of the data center, the total data center information path power consumption is modeled in Equation 40:

where Pinfo is the sum of all powers of each component that is used for carrying information or data in the system. The information path power is the network within the data center (e.g., Prouter, Pswitch, Pfirewall) and other network equipment that carry information. This is similar to the previous definition of signal powers stemming from components on the cascade and it is defined in Equation 41:

and Pnon-info is defined as the power used by the other IT-critical components that process the data but that are not directly involved in data transmission (e.g., Pprocessor, Pmemory, Pstorage, PNIC).This is similar to the previous definition of the non-signal or wasted power of cascaded signal path components. The power consumed by the non-info components is shown in Equation 42:

where NIC represents the network interface cards.

Using this dichotomy to represent a data center in terms of a fine-grain consideration of components, Equation 40 may be used to recast PUE, as defined in Equation 39, as Equation 43:

From Equation 39 and Equation 43, the data center’s total IT power consumption (e.g., analogous to total power consumed by the cascade) can be rewritten in terms of PUE and the useful and wasted powers on the signal path as Equation 44:

Now, just as in previous sections, based on Equations 14, 43 and 44, the waste factor of the data center (ignoring auxiliary power similar to ignoring off-path power) can be defined as Equation 45:

The total power consumption for the data center, like the approach in Equation 9, can then be calculated by considering the data center as a single system that has signal path components, with some that carry information and some that do not, as well as auxiliary non-path components. Using Equation 45, Equation 9 and the definition of PUE, the total consumed power is defined in Equation 46:

The interpretation for W in Equation 46 is intuitive as it is for circuits or communication systems and relates W to PUE, an existing FoM in data centers. However, this interpretation requires a finer definition of components that map to Figure 1, and this approach enables a better understanding of the power efficiency of the data transport.

Figure 6

Figure 6 Illustrative comparison of two data centers.

The example in Figure 6 shows how waste figure applied to a data center provides a more detailed understanding of power efficiency compared to the commonly utilized PUE metric in data center evaluations. Consider two data centers with equal PUE values but with different architectures.

The example of Figure 6 assumes that Data Center A is a larger facility with more equipment and a higher total energy consumption. Conversely, Data Center B is smaller and uses less total energy. Assuming both data centers have identical PUEs and comparing total energy use, it might seem that Data Center B is more efficient. However, PUE, like waste factor, is designed to determine relative or proportionate energy efficiency without respect to actual consumption levels.8,9 Waste factor, with its focus on power wasted on the path that transfers data, provides a better measure of the power efficiency of these two data centers since their ultimate mission is to transfer data in a network. Table 5 shows a comparison of the power consumption on PUE for both data centers.


For Data Center A, the power allocation is as follows: Pinfo,A = 140 kWh for information transmission, Pnon-info,A = 40 kWh for non-data transmission components and Paux,A = 150 kWh for auxiliary equipment. In comparison, consider Data Center B, which allocates Pinfo,B = 60 kWh of power fort5.jpginformation transmission components, Pnon-info,B = 30 kWh for non-data transmission components and Paux,B = 75 kWh for auxiliary equipment. This example has been specifically chosen to ensure the PUE values for each data center are identical (e.g., PUE would indicate they are equally energy efficient).

where PUEA and PUEB denote the PUE of Data Center A and Data Center B, respectively.

Now, using Equation 45, W for the data centers can be calculated:

By evaluating the waste factors, WA and WB, it is apparent that Data Center A is about 20 percent more efficient in its energy use in transporting data, even though both have identical PUEs. This efficiency is measured by comparing the amount of power used directly for carrying and processing information in the system to the overall power consumption. The PUE metric, although standard in the industry, does not capture the detailed energy usage of specific equipment and their relative power waste along the signal path, which is the ultimate job of a data center. It is worth noting that this result is somewhat analogous in Reference 3 where wider bandwidth THz channels are more power efficient on a per bit basis than narrower millimeter wave channels. The waste factor offers a more detailed perspective by considering the function and efficiency of individual components, which may be defined with their own W values. This analysis demonstrates the potential benefits of adopting the waste factor as a metric for evaluating energy efficiency in complex infrastructures such as data centers. Utilizing W as an efficiency metric can provide new insights into optimizing power consumption across various systems involving a source and a sink.

CONCLUSIONS AND FUTURE DIRECTIONS

The waste factor (W) or waste figure (WF in dB) is a new figure of merit for quantifying power efficiency and offers a useful advancement in the field of electrical engineering and system design. By providing a standardized metric for power consumption, it becomes possible for designers and researchers to have a common approach to quantifying power efficiency. This article has shown how the mathematical formulation is similar to the historical approach used to create the noise factor (F), yet W has much broader applications to systems of all types. The waste factor enables electrical engineers and circuit designers to quantify and minimize power waste in any circuity or cascade of devices or systems. This paper has demonstrated the foundational principles of the waste factor, its mathematical derivation and its practical applications across various scenarios, including passive devices, wireless channels, communication systems and data centers. More applications are possible.

Waste factor provides an intuitive understanding and mathematical formulation of power consumption within cascaded systems and allows for the optimization of designs in a manner that was previously challenging due to the lack of a unified metric. Moreover, the application of waste factor in emerging and critical areas such as UAV cellular infrastructure, millimeter wave wireless networks and data centers underscores its versatility and relevance in contemporary engineering challenges. As the demand for energy-efficient solutions continues to grow, the waste factor offers promise as a standard analysis tool for enabling green communications and sustainable technology development. Its adoption as an industry standard could drive significant improvements in the energy efficiency of future electronic devices and systems. The waste factor not only complements existing metrics such as the total PAE in amplifier design or PUE in data center design but also enriches the toolkit available to engineers for designing systems that are not only high-performing but also environmentally responsible. Future work may open new areas of application of the waste factor in both academic research and industry practice with the development of standardized measurement and reporting guidelines and applications of waste factor to other types of systems or devices. Additionally, further exploration of potential applications, such as in the use of AI and ML algorithms for power-efficient design, could lead to more innovative solutions and advancements in energy efficiency.

ACKNOWLEDGMENTS

The authors thank Nicola Piovesan, Antonio De Domenico and Prof. Hamed Rahmani for useful discussions, feedback and suggestions.

References

  1. H. T. Friis, “Noise Figures of Radio Receivers,” Proc. IRE, Vol. 32, No. 7, pp. 419–422, 1944.
  2. Y. Xing, F. Hsieh, A. Ghosh and T. S. Rappaport, “High Altitude Platform Stations (HAPS): Architecture and System Performance,” in 2021 IEEE 93rd Vehicular Technology Conference (VTC-Spring), April 2021, pp. 1–6.
  3. O. Kanhere, H. Poddar, Y. Xing, D. Shakya, S. Ju and T. S. Rappaport, “A Power Efficiency Metric for Comparing Energy Consumption in Future Wireless Networks in the Millimeter Wave and Terahertz Bands,” IEEE Wirel. Commun., Vol. 29, No. 6, pp. 56–63, 2022.
  4. M. Ying, D. Shakya, H. Poddar and T. S. Rappaport, “Waste Factor: A New Metric for Evaluating Power Efficiency in any Cascade,” IEEE Global Communications Conference, Kuala Lumpur, Malaysia, Dec. 2023, pp. 1–6.
  5. T. S. Rappaport, J. N. Murdock and F. Gutierrez, “State of the Art in 60-GHz Integrated Circuits and Systems for Wireless Communications,” Proc. IEEE, Vol. 99, No. 8, pp. 1390–1436, 2011.
  6. J. N. Murdock and T. S. Rappaport, “Consumption factor: A Figure of Merit for Power Consumption and Energy Efficiency in Broadband Wireless Communications,” IEEE GLOBECOM Workshops, pp. 1393–1398, 2011.
  7. J. N. Murdock and T. S. Rappaport, “Consumption Factor and Power-efficiency Factor: A Theory for Evaluating the Energy Efficiency of Cascaded Communication Systems,” IEEE J. Sel. Areas Commun., Vol. 32, No. 2, pp. 221–236, 2014.
  8. M. Dayarathna, Y. Wen and R. Fan, “Data Center Energy Consumption Modeling: A Survey,” IEEE Commun. Surv. Tutorials, Vol. 18, No. 1, pp. 732–794, 2016.
  9. L. A. Barroso and U. Holzle, “The Case for Energy-proportional Computing,” Computer, Vol. 40, No. 12, pp. 33–37, 2007.
  10. S. Lucyszyn, “Power-added Efficiency Errors with RF Power Amplifiers,” International Journal of Electronics, Vol. 82, No. 3, pp. 303–312, 1997.
  11. M. Steer, “Microwave and RF Design V: Amplifiers and Oscillators,” 3rd ed., Vol. 5, NC State University, 2019.
  12. A. A. Sweet, “MIC & MMIC Amplifier and Oscillator Circuit Design,” Artech House, 1990.
  13. L. B. J. Walker, “High-Power GaAs FET Amplifiers.” Artech House, 1993.
  14. T. S. Rappaport, “Waste Figure (W): A Figure of Merit for the Power-efficient (“Green”) Engineering of Any Device or Cascade of Circuits or Systems,“ ITU Webinar, Feb. 20, 2024, https://www.itu.int/en/journal/j-fet/ webinars/20240220/Pages/default.aspx.