Internet quality by location

At a glance

▲now  As of 16:05–16:19 EDT, ending 1 minute ago, 3 of 10 locations were over the 1.0% loss threshold: 630, 629, 125.

▲last 24h  7 location(s) went over the 1.0% threshold at some point in the last 24 hours: 255 (2h), 630 (13h), 629 (13h), 125 (8h). Worst single hour 11.22% at 255.

One block per hour from 17:00 EDT onwards. Green is at or below 1.0% loss, amber above it, orange above 5.0%, red above 10.0%, grey is unmeasured. Hover a block for its figure.
Independent view is the same window measured from a machine on an unrelated datacentre connection, which pings each location from the public internet instead of asking that location's own equipment to send the test. It shares none of the main monitor's machinery, so it is a running check on it. A slightly higher figure there is expected, because that path is longer. A materially LOWER figure would mean the loss is being produced at or near the measurement rather than on the line.

Current conditions — 16:05–16:19 EDT

▲degraded  As of 16:05–16:19 EDT, ending 1 minute ago, 3 of 10 locations were over the 1.0% loss threshold: 630, 629, 125.

The most recent measured minutes on record. This answers "what is it doing now", as distinct from the measurement period summarised below.

Measurement runs continuously, one sample per second per location, and this page is republished every 5 minutes. Page built 16:20 EDT. Newest measured minute on record: 16:19 EDT, 1 minute ago. The measurement period summarised below runs up to 16:20 EDT and is still going - that is its latest sample, less than a minute ago, not an end. This page reloads itself every 10 minutes.

The measurement period — 15:20–16:20 EDT on 2026-10-04

▲degraded  During 15:20–16:20 EDT on 2026-10-04, 4 of 9 locations lost more packets than the threshold allows: 630, 629, 125, 729.

Each location's internet connection is tested once per second by sending a small packet to a public internet address and waiting for the reply. A missing reply means a packet was lost somewhere between that location and the internet.

Measured period: 3600 seconds, one sample per second per location. A single lost second moves a location's loss figure by 0.028 percentage points, so small percentages rest on few events. Graphs below cover a longer window and are labelled separately.

Problems found during 15:20–16:20 EDT on 2026-10-04

Each location over the last 6 hours

One panel per location. The line is the median response time in each slice; rising means slower. The strip underneath marks slices in which packets were lost. Breaks in the line are periods with no measurement, never drawn across. Each panel is scaled to its own range, printed beside its name.

Covers the last 6 hours, 2026-10-04 10:20 to 2026-10-04 16:20 Eastern, across every measurement run in that window. 99.7% of those hours carry data. Measurement is continuous now, so that figure climbs towards 100% as the day goes on; the gaps are earlier periods when measurement ran in short scheduled sessions instead. An unmeasured period is not a quiet period.

Each location during 15:20–16:20 EDT on 2026-10-04

Every interruption during 15:20–16:20 EDT on 2026-10-04

Includes one-second blips, which are harmless alone but indicate an unsteady connection. Times are Eastern.

Terms

clean means loss at or below 1 per cent, watch above 1 and up to 5, degraded above 5 and up to 10, failing above 10. Minutes of data is how many one-minute summaries a current-conditions row is computed from, out of the 15 most recent available - 10 means the figure rests on 600 tests, 1 means it rests on 60. Loss is the share of once-per-second tests that got no reply. Typical ms is the middle response time. Slowest 1 in 20 is the 95th percentile. Worst 5 min is the highest loss over any five consecutive minutes in the period, which catches loss that is concentrated rather than spread evenly. The independent reference is a machine on an unrelated datacentre connection running the same test, so a bad moment at the address being pinged is not mistaken for a fault at these locations. MICRO_DROP is one or two lost seconds, SHORT_FAILURE three to twenty-nine, OUTAGE thirty or more, and SLOWDOWN is at least ten seconds much slower than that location's own median for the measured period - not against its long-run history.