Upstream latency

Median response time of every provider over the last 90 days.

Response time · median per 1 d · times are UTC

Click a line or a row to focus one provider

Median	p95	Failed probes
2.65 s	6.09 s	402 of 8,797
2.38 s	6.54 s	395 of 8,806
2.13 s	10.3 s	1,008 of 16,250
1.73 s	8.00 s	947 of 8,811
1.18 s	8.00 s	1,046 of 8,506
1.07 s	1.69 s	1,405 of 16,250
976 ms	8.00 s	3,884 of 8,801
782 ms	8.00 s	1,770 of 16,250
654 ms	7.54 s	410 of 8,823
650 ms	8.00 s	9,848 of 16,250
343 ms	1.35 s	346 of 16,250
316 ms	2.15 s	215 of 8,803
232 ms	1.27 s	137 of 8,808
63 ms	2.86 s	339 of 8,800
48 ms	694 ms	258 of 16,250
34 ms	931 ms	133 of 8,823
25 ms	321 ms	54 of 8,713

Probes run from the links.et host inside Ethiopia, roughly every five minutes. Each line is the median of the probes in one bucket, over the default egress route only. The amber hatch starts where a response counts as slow and the red one where the probe gives up, both for the provider in focus, or for most providers when none is. A failed probe counts with the time it took to fail, so a provider that times out reads as slow here and one that refuses at once reads as fast. The arrow at the end of a row opens the page for that provider, with its uptime.