DC-to-DC links: link mode, diverse paths, third-party circuits #12

Open
opened 2026-11-17 21:35:20 +00:00 by jaylouisw · 0 comments
Owner

The router is already topology-agnostic: import the duct/manhole network (or the metro fibre ring) and
A->B is routed along it, because engine/street.js only knows about a graph. What is missing for a
site-to-site circuit is the contract, not the routing.

Ask

A design mode for a point-to-point link between two sites, with protection.

  1. Link mode — two named sites, an optic at each end, no splitter and no ONT semantics. The report
    is a circuit: A -> B, routed length, joints, splices, optical budget, verdict, and what to buy.
  2. Diverse paths — the normal DC-to-DC requirement is an A-leg and a B-leg on different duct.
    Minimum-cost edge-disjoint routing (Suurballe) over the same graph, with a diversity report: the
    two paths, their lengths, and every segment they share with the metres ("diverse except 180 m at
    the building entry"), because "diverse until the last duct" is the real answer and must be stated.
  3. Third-party circuits — a large share of these links are bought, not built: a run type that
    carries a provider, a stated length and a cost, contributes its loss to the budget, and is excluded
    from the cable BOM.

Acceptance

  • POST /api/design/link (or synthesise with mode: "link"): two sites, a network, optics; returns
    the path(s), per-leg budget verdicts, and a BOM that contains cable only for owned segments.
  • With a ring graph and two diverse legs: paths[0] and paths[1] are edge-disjoint on the true ring
    and share 0 m; with a graph where diversity is impossible, the result says which segment is shared
    and how many metres.
  • A link whose only route exceeds the optic's reach is infeasible, naming the reach and the measured
    length — same rule as budget.pointToPoint today.
  • Third-party segments appear in the loss breakdown and in the cost, and never in the cable metres.
  • Determinism: same input, same legs.

Not in this issue

Amplifiers, regenerators, CWDM/DWDM multi-wavelength budgets, and duct capacity per segment — each is
its own model and its own decision.

Today

  • routing: works (street.buildGraph + routePath) on any imported geometry;
  • budget: works (budget.pointToPoint), including reach and overload — a short duct run on a ZX optic
    fails on overload and needs an attenuator;
  • auto-design can represent A->B as head-end + one far end, but labels the far end an ONT and ranks it
    as customer service, and only ever produces one path.
The router is already topology-agnostic: import the duct/manhole network (or the metro fibre ring) and A->B is routed along it, because `engine/street.js` only knows about a graph. What is missing for a site-to-site circuit is the *contract*, not the routing. ## Ask A design mode for a point-to-point link between two sites, with protection. 1. **Link mode** — two named sites, an optic at each end, no splitter and no ONT semantics. The report is a circuit: A -> B, routed length, joints, splices, optical budget, verdict, and what to buy. 2. **Diverse paths** — the normal DC-to-DC requirement is an A-leg and a B-leg on different duct. Minimum-cost edge-disjoint routing (Suurballe) over the same graph, with a **diversity report**: the two paths, their lengths, and every segment they share with the metres ("diverse except 180 m at the building entry"), because "diverse until the last duct" is the real answer and must be stated. 3. **Third-party circuits** — a large share of these links are bought, not built: a run type that carries a provider, a stated length and a cost, contributes its loss to the budget, and is excluded from the cable BOM. ## Acceptance - `POST /api/design/link` (or `synthesise` with `mode: "link"`): two sites, a network, optics; returns the path(s), per-leg budget verdicts, and a BOM that contains cable only for owned segments. - With a ring graph and two diverse legs: `paths[0]` and `paths[1]` are edge-disjoint on the true ring and share 0 m; with a graph where diversity is impossible, the result says which segment is shared and how many metres. - A link whose only route exceeds the optic's reach is infeasible, naming the reach and the measured length — same rule as `budget.pointToPoint` today. - Third-party segments appear in the loss breakdown and in the cost, and never in the cable metres. - Determinism: same input, same legs. ## Not in this issue Amplifiers, regenerators, CWDM/DWDM multi-wavelength budgets, and duct capacity per segment — each is its own model and its own decision. ## Today - routing: works (`street.buildGraph` + `routePath`) on any imported geometry; - budget: works (`budget.pointToPoint`), including reach and overload — a short duct run on a ZX optic fails on overload and needs an attenuator; - auto-design can represent A->B as head-end + one far end, but labels the far end an ONT and ranks it as customer service, and only ever produces one path.
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/glassforge#12
No description provided.