Dynamic day-to-day operations sometimes require that data center operators deploy additional equipment without taking time to validate the overall capacity available from circuit breakers. Although adequate processes may be in place, overloaded breakers always seem to fail at the worst possible instant -for example, when an outage affects main power circuits, moving all loads onto backup circuits.

New energy monitoring tools such as Arch Rock Energy Optimizer let users set thresholds on electrical circuits to generate an alert, typically before the power supply has exceeded its “nameplate” rating and well before the actual capacity of a breaker is reached. Different thresholds can be selected depending on the organization’s “safety tolerance.” Figure 1 shows examples of possible thresholds set on the “Apparent Power” metric:

• A 63-amp circuit on 400 volts monitored for 50%, 80% or 100% load
  • 12.6 kVA is equivalent to 50% load
  • 20.2 kVA is equivalent to 80% load
  • 25.2 kVA is equivalent to 100% load
• A 32-amp circuit on 230V monitored for 50%, 80% or 100% load
  • 3.68kVA is equivalent to 50% load
  • 5.88kVA is equivalent to 80% load
  • 7.36kVA is equivalent to 100% load

Figure 1 - Setting a threshold on the Apparent Power metric (e.g. 32-amp circuit at 80% load)

Various energy and environmental metrics, monitored in real time and over extended periods, can be of enormous help to data center operators in rapidly identifying familiar incidents that affect data center performance. For example, new tools available today can record the impact of a partial power outage on temperatures and power metrics.

Figure 1 shows a partial power failure and its direct impact on temperature change. Note that the partial power outage has led to a drop in power demand for the monitored electric load. Figure 2 shows how cold aisle (inlet) and hot aisle (exhaust) temperatures start equalizing due to a partial power failure of the cooling system. In Figure 3, the same root cause generates a quick temperature rise in the IT server room, generating threshold alerts.

Operators can leverage this correlation and analysis of metrics to adjust their thresholds and get quickly alerted to a set of events that they have identified as a partial power outage. When such events occur, they can immediately contact maintenance services to correct the root cause and/or properly shut down the servers, avoiding even more serious damage.

Figure-1: Temperature versus Power Outage

Figure-2: Watch the lab server inlet and exhaust temperatures equalize

Figure-3: Watch the IT server room temperatures rise