Edge Functions

A minute counts as failed only when two or more regions failed the check; the percentages are the minutes that did not. Failures in one region are counted apart: our checks saw them on 66 of these 90 days, 708 minutes in all, and a day with 30 or more such minutes is a disrupted day.

• 10 Nov12 minutes of failed checks; 2 minutes failing in one regionDetected by our checks. No incident was published.
• 13 Oct5 minutes of failed checks; 1 minute failing in one region; some checks hit the 6-second timeoutDetected by our checks
• 22 Sep2 minutes of failed checks; 51 minutes failing in one regionDetected by our checks
• 19 Sep73 minutes of failed checks; 1 minute failing in one regionDetected by our checks
• 1 Sep60 minutes failing in one region; the slowest 5% of checks in the day’s last hour took 2,316 ms, against a usual 958 msDetected by our checks
• 31 Aug37 minutes failing in one regionDetected by our checks
• 20 Aug44 minutes failing in one regionDetected by our checks
• 19 Aug49 minutes failing in one regionDetected by our checks

Slowest 5% of checks (p95)Typical check (p50)Flagged above: twice the usual time

Before 2 Nov, each day’s figures come from its last hour or two only. A day is flagged as slow only when its slowest 5% of checks take more than twice their usual time (the median of the previous 30 days). 2 days were above that line in the last 90 days. Above the chart: on 13 Oct some checks hit the 6-second timeout.

p95 today so far

978 ms

p50 today so far

672 ms

Usual p95 (30-day median)

1,012 ms

Flagged above today

2,023 ms