Doubling the load on a link quadruples the cost — which is why congestion is the rare operations problem that is naturally quadratic, and why routing flows around a hot edge maps cleanly onto an Ising model with detour penalties that keep answers operational.
Quadratic congestion cost means the worst link dominates the bill. Our demonstration finds the move-one-flow optimum — cost 19 versus 20 — at 4.3× the guessing rate.
Rerouting everything is its own outage. Detour penalties inside the same energy landscape keep the sampler honest about operational reality.
Where to add capacity is congestion routing run backwards; the formulation is reused, not rebuilt.
One region of your mesh, real traffic matrix, anonymised — encoded as flows on an Ising landscape with your detour costs, not textbook ones.
QAOA training you can watch converge, sample distributions audited against your current routing table's cost.
Mesh sizes that work today, what the scaling curve promises, and what your traffic-engineering team should take from the formulation — pursue, park, or drop.
These programs are published in our algorithm library. The numbers below come from recorded executions we can reproduce on demand.
Congestion-aware flow routing on a six-node mesh.
Measured: Found the move-one-flow optimum (cost 19 vs 20) at 4.3× the guessing rate.
Results are from the library items' own recorded runs on our simulator — the same one your browser uses.
A backbone routes millions of flows; six nodes is a demonstration. The point is the shape of the problem — naturally quadratic — which makes telecom one of the most honest quantum fits there is.
A region, a traffic matrix, the link that hurts — an engineer replies with the mapping.