Design Comparison
Hold the project brief still, change the system package, then vary one shared assumption at a time to inspect consequences without manufacturing a winner.
Hold one project brief constant
The interactive workbench begins with a 720 kW, twenty-four-rack reference hall, a user-entered energy tariff and a normal inspection state. Change these inputs once and the comparison layer sends the same values to every model that can honestly accept them. IT load enters the legacy-hall thermal model, the liquid-pod heat balance and the electrical-path model directly. The rack model derives an exact per-server distribution that reconciles to the same total load. The generic fire-sequence model has no load or rack input, so the mapping ledger labels both values outside its boundary rather than manufacturing an effect.
A path-A or shared-boundary drill maps only to matching electrical and network nodes. It does not silently invent a pump failure, water-system impairment or other unrelated event. Every exact, derived and outside-model mapping remains visible beside the comparison.
Compare complete reference packages
Air-first retrofit combines hot/cold aisles, conventional air cooling, a single electrical chain, end-of-row access and a generic wet-pipe sequence. Contained shared A/B adds containment, rear-door heat capture, dual downstream electrical lanes with shared dependencies, top-of-rack leaf-spine fabrics and a single-interlock preaction sequence. Liquid-ready independent uses direct-to-chip capture, independent modeled electrical lanes, physically diverse dual-fabric assumptions and a separate double-interlock plus supplemental-agent sequence.
These are deliberately legible teaching configurations, not product recommendations. Each package runs the legacy-room, liquid-pod, power-path, rack/fabric and fire-sequence evaluators. The consequence matrix keeps facility energy, heat path, electrical service, rack fit, packet reachability, protection sequence and the narrow entered energy-cost calculation in separate rows. Red or amber states expose a modeled constraint or dependency; they do not produce a verdict.
See which assumption moves the result
The range explorer evaluates IT load, rack count and the entered energy tariff one at a time while every other brief value and each package definition stays fixed. Choose facility energy, annual energy cost or average rack density, then set a symmetric five-to-fifty-percent low/high range. Centered bands keep the low and high cases distinct, signed labels expose direction and magnitude, exact low/base/high values remain readable, and any modeled consequence-state transition is named.
Factors are ordered only by movement in the selected output. The range is user-selected, not a probability distribution, confidence interval, forecast or Monte Carlo result. A large movement does not make a package better or worse, and zero movement can simply mean that the selected model does not use that assumption.
A dossier that preserves the reasoning
The browser can export Markdown for human review or JSON for later reuse. Both formats include the normalized brief, mapping ledger, complete model inputs, model-version identifiers, package stack, consequence register, current sensitivity observations, unresolved items and shared boundaries. No data is uploaded. The files contain no composite score, package ranking or automatic winner.
Use the output as an educational decision record: state which consequences are acceptable for the named context, identify which assumptions need evidence and assign the specialist validation still required. It is not a bill of material, construction design, code review, certification or commissioning result.
What remains outside the comparison
The workbench does not calculate CFD, hydraulic pressure drop, equipment curves, electrical protection or arc flash, generator dynamics, structural point loads, cable loss budgets, protocol convergence, adopted-code compliance, water, embodied carbon, capital cost, demand charges, schedule or regional availability. Fire choices remain generic cause/effect illustrations pending site-specific hazard analysis, qualified engineering and authority review.
Return to Design Pro to open deeper editable scenarios, or inspect individual Playbook patterns and their evidence records.