Every endpoint as one picture: a row per endpoint, a column per hour of the last day, an aggregate row across the whole marketplace, and a filter to narrow it down. Every figure comes from measurements of real traffic and periodic health probes — nothing here is self-reported, and the board updates live as measurements land.
clean hoursome errorselevated errorsno measurementsHours are UTC, oldest first · refreshes about every 30 s
Response is the median time from request to last byte — the 30-minute window when it has samples, else the 1-day window, labelled either way; every endpoint has one, whether or not it streams. Uptime is the share of successful requests and probes in each window, and the count beside it is the attempts it was measured over, so a measured figure can be told from a lucky one. A provider's uptime weights each of its endpoints by that count, which makes it the uptime of the provider's traffic rather than the mean of its endpoints. An em dash means the window has no measurements yet.