Loading the latest observations…

Outbound bytes from the OCI relay, measured by OCI Monitoring across every network interface in the account, against the Always Free allowance of 10 TB a month. Months and days are UTC, so a day starts at 09:00 JST. The projection adds the 7-day and the last-24-hour pace to what has been used, and the dashed line is the daily pace that would use the whole allowance. This is a measurement, not the bill.

Viewer-facing SLOs are judged against their target with an error budget. A row is not judged when its evidence coverage is below the gate or its sources disagree, and collecting until its 7-day window lies fully inside the measured period; meanwhile it shows the last 24 hours. Quality rows are references, and guardrail rows are constraints and invariants, not availability. Bar: window value; ring: last 24 hours; line: target. Each bar starts at 90% unless a value is lower. Targets are internal engineering objectives, not guarantees. Click a row for coverage, interruptions and the remaining error budget.

The broadcast shows a JST and a UTC clock. Independent observers read both with OCR, check that they agree, and confirm they keep moving at three points along the path.

Strips use 10-minute buckets: green = clock moving, red/amber = a frozen or black picture was seen, grey = the observer could not read the clock, light = standby. Grey is a gap in the check (an unclear OCR read or no sample), not a stream fault; only red/amber mark a problem with the picture. Clock delay is the age of the on-screen clock when the frame is decoded, so it includes encoder and decoder buffering, not only the network. The checks confirm the picture keeps moving; they do not measure what viewers receive.