Loading...
How the platform analyzes geographic regions — footprint, external connectivity, failure impact, resilience, internal structure, and comparison.
A region is defined by a set of landing point IDs. The analysis builds a RegionContext by resolving those LPs to countries, cables, and graph edges. Edges are classified as internal (both endpoints in the region) or external (one endpoint inside, one outside). The analysis uses the country-level topology graph, computed asynchronously in a Web Worker with memoization caches.
Pure counting and sorting on the resolved LP/cable/country sets. No pathfinding involved — computed synchronously and instantly.
How the region connects to the outside world: gateway countries (internal countries on external edges), gateway cables (cables crossing the region boundary), and the ratio of internal vs. external edges.
Iterates over all external edges (graph edges where one country is in the region and one is not). For each, records the gateway country, gateway cable, and the external countries reachable via that edge. Gateway countries and cables are sorted by their external country reach.
Simulates failure of each individual cable and each individual landing point:
First builds a baseline reachability snapshot: for each internal country, BFS finds all reachable external countries. Then for each cable/LP:
LP failure edge resolution: a LP failure only severs a cable-country edge if the LP is the sole LP for that cable in that country. Multiple LPs in the same country provide implicit redundancy.
Criticality scoring:
N+2 failure analysis for regions, probabilistic failure models, correlated failure scenarios (e.g., earthquake affecting multiple cables in the same corridor).
External reachability: For each external country, checks if ANY internal country can reach it (OR semantics). Finds shortest hop count and up to 5 paths per internal source, keeping the top 10 deduplicated paths by hop count.
Transit dependencies: Extracts transit countries from paths. For each, simulates blocking and counts how many external countries become unreachable. Top 10 by affected count.
Redundancy score:
Per-internal-country reachability breakdown, normalized resilience score (0–100), capacity-aware reachability, scenario comparison framework.
Hub countries:Count each internal country's degree in the graph, split into internal neighbors (also in region) and external neighbors.
Components: BFS flood-fill restricted to internal edges only. Countries not connected to each other via internal edges are in separate components.
Fragile pairs: For every pair of internal countries, find internal-only paths (all intermediate nodes also internal). Sort by ascending path count — pairs with count=1 are most fragile (a single edge failure disconnects them).
Fragile pairs limited to top 10and use max 5 paths per pair. A pair that appears robust with 5 paths might actually have only 1 truly independent path. Component analysis is binary (connected or not) — doesn't measure weak connectivity.
Benchmarks the region against all default regions (predefined geographic regions) with rankings and delta-from-median metrics.
For every default region, a full analysis is computed (or loaded from pre-computed JSON). The current region is ranked (1-indexed, descending) on: LP count, cable count, external country count, redundancy score, and critical cable count. For each metric, the median across all benchmark regions is computed and the delta is shown.
Include user-created regions in comparison, percentile-based ranking, multi-metric similarity scoring (find "most similar" regions).
A lightweight preview shown on hover in the region list — profile label, risk tone, and key insights — without the cost of a full Web Worker analysis.
Uses only buildRegionContext, computeRegionFootprint, and computeRegionExternalConnectivity — these are pure set/map operations with no BFS pathfinding. Profile classification:
Risk tone:
All thresholds are heuristic design choices, not derived from network theory. The 50% single-cable LP threshold for "high" risk is arbitrary. A region with 3 gateway cables but all running through the same physical corridor is rated "low" risk.
Full region analysis runs in a dedicated Web Worker. The country graph is built once at worker initialization and reused for all requests. All pathfinding calls are wrapped with memoization caches (path cache, reachability cache, hop count cache, blocked LP edge cache) keyed by serialized parameters.
The region detail page shows an instant overview (context + footprint + external connectivity, computed synchronously via useMemo) while the full analysis loads asynchronously in the worker. The comparison tab is loaded on-demand when the user opens it.
The memoization cache key is a serialized parameter string — cache misses are expensive. Large regions (many LPs, many cables) can produce combinatorially large search spaces despite the bounded pathfinding.