All webinars

Optimization with RCI, RTI, and PUE

How RTI, RCI and PUE guide a four-step CFD optimization of a hot aisle contained data hall, from oversupply to sealed racks.

This webinar applies three metrics, Rack Cooling Index (RCI), Return Temperature Index (RTI) and Power Usage Effectiveness (PUE), to find the best operating point of a data hall. It is presented with a guest co-presenter from the Department of Energy’s data center energy practitioner training, and is aimed at engineers who use computational fluid dynamics (CFD) to tune airflow and energy.

What the webinar covers

The opening defines air management. Recirculation is rack exhaust re-entering rack intakes. Bypass is supply air that passes the racks and returns cold, which lowers the temperature difference across the cooling units. The presenters explain why they aim for the highest workable supply temperature with the least air: better chiller efficiency, lower fan energy, and more hours of economizer use.

Each metric is then explained. RTI compares the air the racks need with the air supplied. RCI measures how many rack intakes fall outside the ASHRAE recommended range and by how much. PUE is total facility energy divided by IT energy. The presenters also discuss fan power rising with the cube of airflow and the dependence of chiller efficiency on return temperature.

The example is a non-raised-floor, hot aisle contained hall cooled by fan walls on both sides, with a ceiling return plenum. Four CoolSim cases are compared:

  1. The base case oversupplies air. PUE is about 2.0, RTI is about 79, and many intakes are colder than needed.
  2. Supply airflow is reduced, and PUE falls to about 1.9.
  3. Supply temperature is raised. Energy improves, but leakage between racks creates hot spots and lowers RCI high.
  4. Leakage between racks is reduced from a very open setting to 10%, which restores compliance with RTI and RCI near their targets.

The talk closes with questions about modeling gaps in a rack, why PUE responds weakly to leakage repairs, and the risks of direct free cooling.

Key points

  • Optimize in order: match airflow to rack demand, fix air management, then raise supply temperature until RCI starts to drop.
  • The presenters use about 125 cubic feet per minute (CFM) per kilowatt as a starting airflow, noting that newer servers often need nearer 100.
  • They cite a rule of thumb of roughly 3.5% better chiller efficiency for each 1.8 °F rise in supply temperature, and describe it as approximate.
  • The guest co-presenter argues that server fans speeding up at higher inlet temperatures costs far less than the chiller savings.
  • RCI counts both the quantity and the magnitude of overtemperature, which a simple count of hot spots does not.
  • Leakage between racks, under racks, and inside racks can undo a good airflow balance. Sealing is a prerequisite to other efficiency measures.
  • PUE changed little between cases because IT energy sits in both numerator and denominator. The presenters prefer a metric that compares cooling energy directly with IT energy.