| Circuit | Down ↓ | Duration | Sides |
|---|
Select an event to review.
Backbone availability
The figure at the top of the page is the average availability of all Points-of-Presence (POPs) over the selected window.
Only POPs with availability data are included. When some POPs have partial history, their clipped availability percentage is used in the average.
A POP typically has two or more routers providing redundancy. The POP is only considered down when every one of its routers is simultaneously fully down or when it has no remaining WAN/backbone connection to another POP.
If one router fails but another stays up, and at least one POP-to-POP backbone connection remains available, the POP remains available — traffic continues to flow through the surviving path.
POP Availability % = (window duration − time the POP was unavailable) ÷ window × 100
Each router maintains BGP sessions with three Route Reflectors (RRs) — core routers that distribute routing information across the backbone. The availability calculation monitors BGP session up/down events recorded in each router's syslog.
Router Availability % = (window duration − time fully down) ÷ window × 100
Router reboots are detected from the boot time reported by show version. When a router's observed boot time moves forward, the reboot appears in the Router Reboots review tab.
Reviewed reboot events marked as Degraded Only or POP Outage are included in router availability. Events left as Needs Review, marked No Customer Impact, or ignored do not affect availability.
The reboot unavailable interval starts at the observed boot time. Recovery time is inferred from the first available signal after that boot: a route-reflector BGP UP event, a current route-reflector Established timestamp, or an ISIS UP event. If none of those are available, the collection time that detected the reboot is used as an estimated recovery time.
POP availability uses overlap. If two routers in the same POP reboot at the same time, the POP is counted down only for the time when all POP routers are simultaneously unavailable.
If planned reboots are staggered so at least one POP router remains available, the individual routers may show degraded availability, but the POP remains up.
When a router has been in service for less time than the selected window, its availability is measured only from its earliest known boot — not from the start of the window. Showing 100% for a period before the router existed would be misleading.
When a POP has less history than the selected window, its row says since and the date its data starts. The downtime strip is grey before that date.
Each backbone link runs IS-IS, a routing protocol that tracks adjacency state between neighbouring routers. When a link fails, IS-IS logs an adjacency-down event; when it recovers, it logs adjacency-up. These events are collected from each router's syslog on every poll.
Only outages confirmed during manual review are counted against uptime — unreviewed events are excluded so noise does not affect the numbers.
Circuit Uptime % = (window − confirmed outage seconds) ÷ window × 100
These metrics come from independent third-party probe measurements, pulled from Grafana. They are averaged over the selected window and are entirely separate from the IS-IS and BGP availability data above.
The POP list shows a fixed 24h summary for latency and jitter so it stays responsive. Click a POP row to see each circuit. Those expanded circuit calculations are loaded on demand because latency and jitter queries are more expensive than the availability calculations.