Skip to main content
01Data Center Cooling Optimization

How do I raise rack power density without guessing at airflow and losing redundancy margin?

Rack-to-row-to-room airflow simulation for high-density and liquid-cooled data center design.
02The problem

A server passes thermal qualification in isolation on the bench, and then throttles in production because neighbouring racks in the row change the inlet temperature and recirculation pattern the single-rack test never saw.

As rack power density climbs past what ASHRAE TC 9.9 air-cooled envelopes comfortably support, teams are evaluating rear-door heat exchangers, direct liquid cooling and two-phase immersion without a clear way to compare the options against their actual room layout and CRAC/CRAH capacity.

Why it's hard
  • Rack-level thermal qualification rarely represents the aerodynamic interaction between adjacent racks and rows in a populated data hall
  • Hot-aisle/cold-aisle containment effectiveness is highly sensitive to floor tile placement, cable congestion and small leakage paths that as-built facilities rarely match as-designed
  • Rising power density is pushing designs past ASHRAE TC 9.9 air-cooled limits, and liquid and immersion cooling introduce two-phase heat transfer physics most electronics-cooling workflows have not modelled before
  • CRAC/CRAH capacity planning decisions are made at the facility level while chassis-level airflow decisions are made at the equipment level, and the two are rarely reconciled in one model
  • Redundancy (N+1) scenarios change the airflow pattern in ways that are difficult to predict without modelling a full row or room, not a single rack
03How we solve it
Predict — simulation
  • Model rack and chassis-level airflow and component cooling in Flotherm
  • Scale to row and room-level airflow, containment effectiveness and CRAC/CRAH interaction in STAR-CCM+
  • Resolve two-phase dielectric immersion and direct liquid cooling physics in STAR-CCM+ when air cooling can no longer support the power density
Measure — hardware
  • Characterize server and power-supply component thermal behavior with Simcenter Micred T3Ster to validate the chassis-level model feeding the room simulation
Correlate — close the loop
  • Validate the room-scale CFD model against facility temperature and airflow sensor data where available
  • Use the correlated model to evaluate N+1 redundancy scenarios and liquid-cooling retrofit options before committing capital
05Standards in scope
  • ASHRAE TC 9.9
  • NEBS GR-63-CORE thermal
07Questions engineers ask

Model your data hall before you commit to a cooling retrofit.

Tell us your current rack density, cooling method and where you're seeing thermal margin erode.