This simulator models a common data center dual-path power topology: two independent feeds (Path A and Path B), each terminating in a hot-swap server power supply, both supporting one shared IT load. Adjust demand and per-path capacity, fail a path or a shared dependency, and watch how load sharing, takeover and capacity margin respond.
• Facility laboratory tab: a real-time 3D workbench of the feed A/B switchboards, dual server power supplies and the common server DC load, with Home view, Focus selected part, Show full enclosure/exploded view toggle, Auto rotate, Expand and a clickable component list with callouts; live stats (A path power, B path power, served load, unserved load, surviving capacity margin, IT demand); a "what is happening" operating-sequence readout; and experiment controls — Pause/advance simulation (10 ms or 1 s steps), playback speed (10× slow motion to 1 minute per second), IT load and per-path capacity sliders, Path A/Path B healthy checkboxes, a shared downstream fault checkbox, and action buttons (enable facility, stop experiment, open path A, open path B, fail shared dependency, restore dependencies, reset protection). • Curves & measurements tab: a torque/load operating-point chart showing A/parallel power, B power and capacity margin over time, a load-sharing history chart, the model equations (normal sharing PA = PB = Pload/2, single-path takeover Psurvivor = Pload, required rating Prating ≥ Pload) and snapshot measurements. • Experiments tab: guided scenarios (normal sharing at 80 kW on two 100 kW paths, single-path takeover after opening A, insufficient remaining rating at 160 kW demand with one 100 kW path, and a shared-dependency failure), a model verification bench (Run model checks) that tests independent fresh models — including full takeover, double-failure deficit, common-dependency loss and negative-capacity guards — and a timestamped event log with a trial-report export. • Learn & assess tab: lessons on separate paths, takeover headroom, cascading overload and common dependencies, a knowledge-check quiz with reset, and a scope/references note.
In this model, two electrically independent paths — Feed A and Feed B — each feed their own server power supply, and both supplies support the same downstream IT load. Under normal, healthy conditions the load model splits demand equally between the two paths. When one path is opened (fails), the healthy path must pick up the full load; the module simulates a hot-swap PSU whose surviving output takes over completely, subject to that path's rated capacity.
If the surviving path's rating cannot cover the full demand, the model latches the overloaded path unavailable — resetting protection cannot restore capacity that was never adequate. A separate "shared downstream fault" toggle demonstrates that A/B labeling alone does not eliminate common-mode failures: if both feeds ultimately depend on the same downstream equipment or load node, failing that shared point defeats both otherwise-healthy paths simultaneously.
The load-sharing chart plots A-path power, B-path power and IT demand over simulated time; watch it flatten to an equal 50/50 split under normal operation, then jump to 100% on the surviving path the instant the other is opened. The surviving-capacity-margin metric tells you how much headroom remains after a failure — a healthy margin close to zero signals the surviving path is near its rated limit, and a negative or blocked outcome shows unserved load appearing.
The Run model checks bench builds independent fresh model instances (leaving your current trial untouched) and verifies invariants such as: three 50 kW modules leave 100 kW capacity after one failure, two failures leave a 50 kW deficit, a common-dependency loss defeats spare capacity entirely (served drops to zero), failed-module counts above the installed count cannot yield negative capacity, the A/B architecture's surviving path carries the full load within its rating, and capacity margin correctly follows the count of remaining healthy modules. This is a logical dual-cord load-sharing model — it does not include rectifier droop, load-supply hold-up time, cross-tie switching, AC fault behavior or breaker time-current curves.
It means the IT load is served by two independent power paths (A and B), typically ending in dual-corded server supplies, so a failure on one path does not interrupt supply as long as the surviving path has enough capacity to carry the full load.
No. Under normal conditions each path might carry 80 kW, but if one path fails, the survivor must carry the full 160 kW, which exceeds its 100 kW rating. Full redundancy for a given load requires each path to be rated for the entire load, not just half of it.
No. If both paths share downstream equipment, controls, or a load dependency, a failure at that shared point can defeat both feeds even though each path individually is healthy. The simulator includes a dedicated shared-dependency fault toggle to demonstrate this.
The Run model checks button builds independent fresh model instances and verifies invariants such as post-failure capacity, deficit calculations after multiple failures, common-dependency defeat of both paths, and that capacity margin tracks the correct number of healthy modules — without altering your current trial.