What these numbers tell you.
The chart shows the median of the available per-probe average ping round-trip times for each location selection. The range shows the lowest and highest probe averages, not individual-packet extremes. Most selections have three ping results; Warsaw has two in the latest run because one probe was offline. For two results, the median is their arithmetic mean. An offline probe has no latency result and is excluded from the median and packet denominator. Cyprus is a country selection spanning Limassol, Larnaca and Nicosia; the other selections are cities. Results describe these samples at these times, not a regional average or service guarantee.
Same city. Different paths.
The latest Athens probes averaged 19.762, 85.360 and 114.476 ms. The original snapshot also showed a wide spread: 19.735, 82.454 and 114.514 ms. The original fastest MTR includes cyta-cy.gr-ix.gr; another includes Cogent router names associated with European cities before CYTA. These observations show why the source network matters. Router names are clues, not proof of physical cable routes. Select either run to inspect its own source logs.
Keep packet loss in context.
The latest ping sample received 559 of 560 packets: Dubai M247 received 15 of 16, while the other 34 responding probes received all 16. Warsaw ZINET.NET.PL was offline during ping and is recorded as missing, not as packet loss. All 36 separate MTR traces reached the destination, including Warsaw ZINET.NET.PL. The Nicosia BrainStorm MTR recorded 14 of 16 destination replies (12.5% loss), while its separate ping received all 16. The other 35 MTR destination rows recorded all replies. Reaching the destination does not mean zero packet loss, and intermediate-hop nonresponses are not added to destination loss. The original run remains available: 575/576 ping replies and 35/36 MTR traces reaching the target.
Methodology and test machine.
The latest archived run took place from 07:06:10 to 07:10:35 UTC on 16 September 2026, targeting 213.111.177.231. It reselected probes from the first collection. There were 36 selected ping probes; 35 returned statistics for 16 packets each and one was offline. MTR was a separate set of 36 traces. The original collection ran from 16:07:09 to 16:11:47 UTC on 15 September using Globalping CLI v1.5.2; the repeat archive does not record a CLI version. These are location samples, not a continuous time series or a matched-probe trend. The operator identifies the target as a KVM Basic VPS in Limassol with 1 vCPU, 4 GB RAM, Debian 13 and hostname globalping-test, prepared with no other workload and no artificial bandwidth limit.
- Kernel: 6.12.107+deb13-amd64
- Measurements originate at Globalping probes and target the VPS
- Ping and MTR are separate measurement runs
- A short sample is not an uptime, throughput or application benchmark
- The operator-provided machine details are context; the archives record the target, timestamps and network output. Only the original archive records the CLI version.
Repeat the test from your own network.
Start with the published Looking Glass for ping, traceroute and MTR. For inbound tests, ping the Looking Glass address from your own machine. Compare several source networks and repeat at different times. For deployment decisions, also measure the application protocol and the dependencies your VPS will contact.