Overview

When laying out a data center (IDC) floor, you often need to compare many alternatives while varying parameters such as rack layout and supply air rate. General-purpose CFD is accurate, but the long per-case runtime makes it impractical to run enough alternatives. Boreas is built for this screening stage — it simplifies the mesh and numerics so a single layout can be solved in minutes to tens of minutes.

In practice, this typically plays out in two stages:

1. Layout / airflow screening — quickly compare many cases while varying rack layout and supply air conditions, narrowing down to a promising layout.
2. Verification analysis of the finalized layout — run a precise verification analysis (commercial or open-source CFD) on the narrowed-down layout.

This document is an example of stage 1: a steady-state cooling airflow analysis of an 800-rack virtual data center room, computed with Boreas.

Model

• Rack layout: 40 rows x 20 racks = 800 racks (cold aisle / hot aisle arrangement)
• Total heat load: 5 MW
• Supply air: 440.0 m3/s total (0.925926 m/s per floor tile)

This is not based on an actual building floor plan. It's a simplified demo geometry made by repeating a single rack row, with no architectural constraints such as column spacing or a real room shape. The purpose is to show that compute speed stays practical even at an 800-rack, 5MW scale.

Full geometry — 40 rows x 20 racks, 800 racks total

Steady-State Temperature Distribution

Steady-state temperature distribution across the whole room. Cold aisles (supply side, low temperature) and hot aisles (exhaust side, high temperature) alternate between rows of racks.

Temperature range 16-28°C — blue (cold aisle) and red (hot aisle) alternate

Zooming into the space between rack rows shows the temperature gradient at the cold-aisle/hot-aisle boundary more quantitatively.

Compute time was about 10-15 minutes on a consumer-grade GPU and about 2 hours on an 8-core CPU — fast enough to compare many alternatives during design screening.

Quantitative Results — Rack Inlet/Exhaust Temperature

Inlet (t_in) and exhaust (t_ex) temperatures for the front 12 of the 800 racks (R1-R12), aggregated at steady state.

Inlet temperature (blue) stays even across racks, and the rise to exhaust temperature (orange) is likewise uniform

Inlet temperature stays within the target range across all 12 racks, and the temperature rise across each rack is uniform with no significant rack-to-rack deviation — no local hotspot racks occurred under this layout and supply air condition.

Local Flow Field — Cold Aisle Supply Flow

Streamlines and velocity vectors show supply air rising from the cold-aisle floor tiles and curving into the rack inlet face.

Cold (blue) supply air rises from the tile and enters the rack inlet face; hot (orange/red) exhaust air leaves through the rack into the hot aisle

Velocity magnitude on the same cross-section. Velocity is highest directly above the supply tile and in the narrow gap between racks, while the center of the cold aisle is a relatively stagnant region.

Summary

This example shows the kind of output Boreas produces when used for data center cooling airflow screening.

• Quantitative visualization of temperature and airflow distribution across the room
• Identification of local hotspot racks
• Compute speed (about 10-15 minutes per case) fast enough to compare alternative rack layouts and supply air conditions
• Extraction of rack-level inlet/exhaust temperature and temperature rise (ΔT) as quantitative data for reporting

In an actual project, the layout narrowed down through this screening stage is then re-verified with a higher-fidelity CFD analysis as the second stage of the workflow.