NwTwin - Network Survivability Digital Twin

The scenarios

Every run starts from Base, the network exactly as modeled, and then re-scores it once per failure it is asked to imagine. The gap between the two is the number that matters: a network scoring well on Base and badly on B+2L is one good day away from a bad month.

Scenario	Runs	What it removes	Answers
`Base`	always	Nothing: the network as modeled	The reference score everything else is measured against
`B+1N`	always	Every single node, one at a time	Which node hurts most when it goes, and by how much
`B+NL`	no trails	A node and a link together	Whether a maintenance window overlapping a fault is survivable
`B+1L`	no trails	Every single link, one at a time	Which spans are load-bearing and which are genuinely redundant
`B+2L`	no trails	Every pair of links	The double-failure exposure that single-fault design misses
`B+1T`	trails	Every trail, one at a time, all of its links together	What one cut costs when several links share a path
`B+TL`	trails	A whole trail, plus one more link anywhere	Whether the network survives a cut while something else is already down

Naming a trail changes the set. Base and B+1N always run. After that a network with no trail named gets the link trio, and one that names even a single trail gets the trail pair instead. The two sets are not comparable, so a score before and after your first trail is a different measurement, not a regression.

Scores come back at three levels, the network, each node and each link, for the base network and separately for the failure scenario you have marked. Design quality and current exposure stay side by side in one view.

Trails

Two links between different pairs of nodes look independent on a map and are frequently not: the same duct, the same fiber bundle, the same bridge. A trail is how you say so. Links sharing a trail name ride one physical path, and the simulation fails them together.

Without trails the model quietly assumes every link fails on its own, which is the assumption that makes a network look survivable right up until the backhoe. Naming trails is the single highest-value thing you can add to a topology that is otherwise already drawn.

Naming one

Double-click a link on the map and fill in its trail field. Any two links carrying the same name are on the same trail; the name is yours, so duct-7 or river-crossing is as good as T1. A link that rides more than one shared path takes a comma-separated list:

trail: duct-7
trail: duct-7,bridge-north

A link left blank is not excluded from the run. It becomes a trail of its own, one link long, so B+1T still fails it individually the way B+1L would have. Naming the trails you know about therefore costs you nothing on the links you do not.

Using them well

Model the shared thing, not the pretty one: a trail should be a duct, a cable, a conduit or a crossing, something a single event can take out. Grouping links because they are in the same city is not a trail, it is a map color.
Start with the crossings: rivers, railways, bridges and tunnels are where the most links share the least ground, and they are usually the first thing B+1T lights up.
Do not name a trail per link: singleton names add nothing that a blank field does not already do, and they make the trail list harder to read.
Re-run after naming: scores are stored, so a network already marked Ready keeps its old numbers until you run it again.
Compare like with like: because the scenario set swaps, judge a change by re-running both versions, not by holding a pre-trail score next to a post-trail one.

The useful question a trail answers. Not "is this link protected", which a map already tells you, but "how much of my redundancy is imaginary". Two diverse-looking paths in one duct score as one path, and that is the number worth knowing before you buy the third.

What it tells you to do

After a run, the twin names the specific actions worth taking:

Add this link: the new link that raises the index the most, with the percentage it gains and the score it reaches
Retire this one: links that can go with no loss of survivability, or the one that costs least if none are free
Restore this first: in the current outage, the link whose recovery returns the most survivability
Ask follow-ups: the AI wizard holds the full picture, so “why?” and “what if we did both?” are ordinary questions

All of it is arithmetic on your own topology, computed locally. The app runs the scenarios, and an assistant can drive the same run through the MCP tools if you would rather ask than click.