The power of power measurement


Electrical power consumption by a computer cupboard, IT room, tech suite, data centre or facility is one of my favourite [pet!] metrics for several reasons:

  • It is readily measured using a wattmeter, watt-hour meter or ammeter on the main supply line/s;

  • Compared to more technical metrics, power is simple to plot, report, explain and understand;

  • As the installed IT equipment and usage gradually changes, so does the power consumption. It is straightforward to track and predict the overall trends without necessarily measuring and controlling every single item and change; 

  • Step changes in power consumption indicate substantial changes in the IT equipment or usage. Marked decreases are welcome but quite rare (e.g. as older equipment is retired from service or replaced by more modern, energy-efficient stuff), whereas marked increases in consumption - especially if unexpected - may be cause for concern;

  • The first law of thermodynamics tells us that all the input energy has to go somewhere i.e. heat which can be costly to remove, increases global warming, increases fire risks and decreases equipment lifetimes. 

In more detail, a high PRAGMATIC score (~77%) indicates that IT power consumption is a valuable metric, well worth considering:

  • Predictiveness: 50%. Power consumption lags changes in the equipment and usage, but gradual and step-changes in consumption are useful for planning updates, responding to adverse trends and demonstrating improvements;

  • Relevance: 75%. The metric is relevant to several important aspects of business including costs, physical security, environmental impacts and sustainability, change management and information security;
        
  • Actionability: 75%. It is quite straightforward to investigate and address worrying trends or unexpected changes in power consumption. Further information may be required to identify causes and determine responses however; 

  • Genuineness: 95%. Provided the measuring equipment itself is capable, properly engineered, installed and used, power measurements are strongly linked to actual consumption;
      
  • Meaningfulness: 90%. Even a basic appreciation of physics is sufficient to understand the metric and its implications. It is one of the simplest measures. Aside from infosec, it is clearly also relevant to sustainability;

  • Accuracy: 95%. Calibrated measuring equipment is readily available if required, while even uncalibrated meters should indicate trends and step-changes sufficiently accurately for management purposes;

  • Timeliness: 35%. Although this is a lagging metric, the lag is short. Power issues should become evident shortly after they occur;

  • Integrity: 85%. It would be quite hard (and potentially dangerous) for someone to manipulate the data e.g. by meddling with the measuring instruments and processes. While electrical/maintenance engineers could design or amend the system to undermine the measurements, electrical safety regulations and inspections reduce the risk, while the motivation for such fakery is low. It is in their interest to ensure that the measurements are reasonably accurate since they are accountable for the reliability and safety of the power systems*;

  • Cost-effectiveness: 95%. This relatively cheap and easy measure generates a lot of value for the business, and for the planet. For instance, the headroom between IT's power consumption and capacity is one aspect of IT resilience: if the meters are at the top of the scale and the wires are warm or glowing, it's only a matter of time before a catastrophic incident.    

In short, IT power consumption is an excellent metric. I commend it to the house.


* However, I would be wary of using this metric to drive and determine bonuses for the engineers, facilities management and IT professionals without additional controls. Lacking sensible management oversight and control, a crude business objective to reduce IT's net energy consumption could lead to sub-optimal strategies (such as deliberately under-reporting, moving processing off-site, or unduly restricting power-hungry processing regardless of the business impacts) rather than genuine improvements.