Pulse

Two truths for every service: what we measure from the network, and what the vendor announces on its own status page. The interesting column is the gap between them.

Divergence — measured down, officially fine

Country network signals

Routing health — can the internet still see these networks?

Operator-run networks first; the dimmed rows are shared CDN infrastructure, whose routing health is a fact about the CDN rather than about the country.

Service board

Incidents

How this works. Every service is probed in four layers — DNS, TCP, TLS, then HTTP — and the verdict comes from the deepest layer that gave a definitive answer. A WAF answering 403 is recorded as up · walled, not as an outage: the service answered us and declined, which is a completely different fact. Naive checkers get this wrong and report every bank as down.

Official status is read from each vendor's own feed, discovered automatically — Atlassian Statuspage exposes /api/v2/status.json on every instance, which is one adapter for a large part of the software industry. Discovery validates the payload structure rather than trusting a 200, because parked domains happily return 200 for any path you ask them for.

Routing health is measured per network rather than per country. Each service is mapped to the ASN announcing its address — derived from what the prober resolved, not a hand-written list — and we record what fraction of RIPE's ~325 routing collectors still see that network's prefixes. Ogero sits at 325/325 normally; a fall is a transit failure in progress, visible below the level any status page describes and well before an individual website check notices. Note the limit this exposes: most regional banks sit behind Cloudflare or Akamai, so BGP tells you about the CDN and not about them — it is a real signal only for operators who run their own network, which is why those are listed first and the rest are dimmed.

Country signals come from RIPE Atlas (connected probes — real hardware in real buildings), RIPEstat (ASNs still visible in the global routing table) and IODA at Georgia Tech. Three independent methodologies, so no single feed can raise a national alarm on its own.

Incidents need a streak and they need agreement across vantage points before they open. This instance currently probes from one location, which is stated plainly rather than hidden: single-vantage confidence is capped, and geo-filtered services are the reason regional probes matter.