Global Traceroute
Run a traceroute to any host from probes on several continents at the same time and compare the paths hop by hop, with the AS number of every transit network in between.
About Global Traceroute
A traceroute from your own machine describes exactly one path: the one from you. Global Traceroute runs the same measurement from probes on different continents in parallel, so you can see how differently the internet reaches the same destination depending on where the traffic starts. Each result lists every responding hop with its IP address, its reverse-DNS name where one exists, the average round-trip time and, most usefully, the autonomous system number that hop belongs to - which turns a column of addresses into a readable sequence of networks.
The AS column is where the diagnosis usually happens. A path that leaves the probe's access network, crosses one or two transit providers and arrives at your hosting AS is healthy. A path that crosses five different transit networks, or that leaves the continent and comes back, is paying for someone's routing policy rather than for distance. When one region's path is dramatically longer than the others, the cause is almost always a missing peering relationship or a route that is only visible through an expensive upstream, and both are things you can raise with your provider using this output as evidence.
Two features of traceroute are routinely misread. Hops that show no reply are common and usually harmless: many routers are configured not to generate ICMP time-exceeded messages, or rate-limit them heavily, while forwarding your traffic without any trouble at all. Latency that appears to drop at a later hop is equally normal, because each hop's timing is measured independently and depends on how quickly that particular router chooses to answer, not on how fast it forwards. Only the trend towards the destination is meaningful, and only the final hop's latency describes the service.
Routing is also asymmetric far more often than people expect. The path from a probe in Singapore to your server and the path back can traverse entirely different transit networks, so a traceroute in one direction never fully explains a problem; when latency is unexplained, ask for a traceroute back towards the probe as well. Measurements are executed by the Globalping open probe network and orchestrated from our own EU infrastructure, with results cached for ten minutes so that a repeated check costs no measurement and returns immediately.
How to use it
- 1Enter the hostname or IP address you want to trace, for example example.eu.
- 2Select Trace from 4 continents; one traceroute runs from each region simultaneously.
- 3Read each block top to bottom: the probe's city and network, then every hop with its address, AS number and average latency.
- 4Compare the AS sequences between regions - the number of transit networks crossed matters more than the number of hops.
- 5Ignore hops that show no reply, and judge quality from the trend towards the destination rather than from individual hops.
Common use cases
- -Finding out which transit providers carry your traffic from each part of the world.
- -Proving to an upstream that traffic from one region is being routed the long way round instead of via a nearby exchange.
- -Checking whether a new peering arrangement or IX connection has actually shortened the path it was meant to shorten.
- -Investigating regional slowness that latency alone cannot explain, by looking at where the path goes.
- -Documenting the route before and after a network change, so the effect of the change is visible rather than assumed.
Frequently asked questions
- Why do some hops show asterisks instead of a latency figure?
- That router did not send an ICMP time-exceeded reply for the probe. It is usually deliberate: generating those replies is low-priority work, so many operators disable or rate-limit them. A silent hop in the middle of an otherwise complete path is not a fault. What matters is whether the final hop, your destination, answers.
- Why does latency sometimes decrease from one hop to the next?
- Because each hop's round-trip time is an independent measurement of how quickly that specific router answered a probe, not a cumulative total. A busy router can take several milliseconds to generate its reply while forwarding traffic instantly, making the next hop look faster. Read the overall trend and the destination figure, and ignore individual dips.
- The paths from different continents share almost no hops. Is that a problem?
- No, that is exactly what you should expect. Each probe reaches your host through its own access network, its own transit providers and its own interconnection points. Comparing those paths is the point of running the measurement from several places: it shows which regions get a short clean route and which are being carried a long way round.
- What is the difference between this and MTR?
- Traceroute sends a few probes per hop and shows you the path. MTR keeps probing continuously and reports per-hop packet loss and jitter as well, which is what you need to locate where quality degrades rather than merely where packets go. Use traceroute to understand the route, and the Global MTR tool when you are chasing loss or instability.