Traceroute
Trace the route from probes around the world to any host, hop by hop: each router’s address, name and round-trip time, silent hops, and where the delay is added.
How to Use
- Type a hostname such as wikipedia.org or a public IP address.
- Choose how many probes trace it (2 to 6) and where they are. Every continent puts them far apart, so you see different routes.
- Leave the protocol on ICMP (like Windows
tracert), or try UDP (classic Unix traceroute) or TCP to a port such as 443 when a firewall drops the others. - Press Run live traceroute. Each probe’s hop list appears when it finishes, usually in 5 to 15 seconds. Until then the page shows a recorded example.
- Pick a probe to see its hops in the chart and in Show Work: which router answered each TTL, where the time jumped, and which hops stayed silent.
A live run uses 3 of the 250 free Globalping tests per hour for your IP address.
Worked Example
wikipedia.org from Helsinki, Buffalo and Tokyo (ICMP, recorded 6 October 2026). Each probe was given a different address by DNS: 185.15.59.224, 208.80.154.224 and 103.102.166.224. Helsinki reached its target in 9 hops with one silent hop (hop 7) and a final round trip of 29.6 ms; Buffalo needed 10 hops, 3 of them silent, and ended at 16.9 ms; Tokyo took 9 hops with none silent and ended at 75.2 ms.
Where Tokyo’s time went. Hops 1 to 4 answered in 0.24 to 1.24 ms, so all four are within 1.242 ÷ 2 × 299.79 = 186 km of the probe. Hop 5 (ae-4.r25.sngpsi07.sg.bb.gin.ntt.net) answered in 75.3 ms, and every hop after it stayed between 75.2 and 75.3 ms. The jump is 75.342 − 1.242 = 74.1 ms, and it persists to the destination, so it is real path delay. At light speed, 74.1 ms of extra round trip allows 74.1 ÷ 2 × 299.79 = 11,107 km more distance each way: a long-haul link, not a fault.
The common mistake: blaming the hop where the number goes up and then down. From Helsinki, hop 5 took 31.664 ms and hop 6 only 28.247 ms, 3.4 ms less. Hop 6’s packets pass through hop 5, so at least 3.4 ms of hop 5’s time was hop 5 being slow to send its own ICMP reply. Read the trend to the destination (29.6 ms), not one hop.
Private hops are normal. Helsinki’s first hop is 192.168.1.1 and Buffalo passes 10.8.14.145 and 10.8.3.233. All three are RFC 1918 private addresses used inside the probe’s own network.
Show Work
How Traceroute Works
TTL, Time Exceeded and Why Hops Go Silent
The trick. Every IP packet carries a Time to Live field (RFC 791) that each router lowers by one; when it reaches zero the router throws the packet away and sends an ICMP Time Exceeded message back to the sender (RFC 792, type 11). Traceroute sends packets with TTL 1, then 2, then 3, and lists who sent each Time Exceeded message, until the destination itself replies. Van Jacobson wrote the first traceroute at LBL (Lawrence Berkeley Laboratory); his source code is signed 20 December 1988 and credits the idea to Steve Deering. It sent three UDP probes per TTL to an unlikely port, up to 30 hops, with a 5-second wait.
An attempt to replace it. In January 1993 RFC 1393 (G. Malkin) proposed a Traceroute IP option, so one packet could collect the whole path instead of 2n packets for n hops. It was Experimental, was never widely deployed, and RFC 6814 moved it to Historic in November 2012. Every traceroute today still uses the TTL method.
Why a hop shows *. A star means no answer within the timeout. Routers are allowed to limit how many ICMP error messages they send (RFC 1812 §4.3.2.8 for IPv4 routers); IPv6 nodes must rate-limit ICMPv6 errors, and RFC 4443 §2.4 warns that simple timer-based limits do not cope with bursty traffic like traceroute. Firewalls that drop ICMP, or routers configured not to answer, give the same result. Microsoft’s tracert documentation calls such routers invisible; when the hops after a star still answer, traffic is passing through it normally.
Asymmetric paths. Traceroute only discovers the way out. Each hop’s time is a full round trip, and the Time Exceeded message can come back by a different route; RFC 1393 already noted that the method “does not trace the return path, which may differ from the outbound path”. The address shown is usually the router interface that faced the probe. To see the other direction, trace from the far end.
About This Tool
This traceroute runs on the Globalping network, a public set of probes run by jsDelivr, from 2 to 6 probes that you place on every continent, on one continent, or by country, city, network or cloud region. Choose ICMP, UDP or TCP (with a port) and IPv4 or IPv6.
For each probe you get every hop: its TTL, address, reverse-DNS name, each packet’s round trip, the best time and the change from the hop before, plus a note when the address is private, carrier-grade NAT or another special-purpose block from the IANA registry. The chart plots all probes so you can compare routes; Show Work explains the chosen probe hop by hop. Open Raw output for the probe’s own traceroute text.
Privacy: the request goes from your browser to api.globalping.io, not to this site, and only when you press Run. Private, loopback and documentation addresses are refused before anything is sent. Sources: RFC 791, RFC 792, RFC 1393, RFC 1812, RFC 4443, RFC 6814, RFC 1918, RFC 6598 (rfc-editor.org), the IANA special-purpose address registries, the LBL traceroute 1.4a12 source, and Microsoft’s tracert documentation.
Related tools: Ping Test, IP Lookup, and DNS Lookup.
Frequently Asked Questions
What does * * * mean in a traceroute?
No reply came back for that TTL before the timeout. The router still forwarded the traffic, because later hops answered, but it did not send the ICMP Time Exceeded message or the message was filtered on the way back. Routers are allowed to rate-limit the ICMP errors they generate (RFC 1812 §4.3.2.8; for IPv6 RFC 4443 §2.4 requires it). In our example the Buffalo probe had 3 silent hops out of 10 and still reached wikipedia.org.
Why is a later hop faster than an earlier one?
Each hop’s time is a separate round trip that includes the router’s own time to build the ICMP reply and the path that reply takes back, which may differ from the forward path. From Helsinki, hop 5 answered in 31.7 ms but hop 6 in 28.2 ms. Only a delay that carries on to every later hop and the destination is a real delay on the path.
Why do I see private addresses like 192.168.1.1 or 10.x in the path?
Routers inside a home or a provider’s network often use private addresses (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16, RFC 1918) or carrier-grade NAT space (100.64.0.0/10, RFC 6598) on their internal links. They can still send Time Exceeded messages, so they show up as hops. The Helsinki probe’s first hop was 192.168.1.1 and the Buffalo probe passed 10.8.14.145 and 10.8.3.233.
What is the difference between traceroute and tracert?
Windows tracert sends ICMP Echo Requests with TTL 1, 2, 3 and so on, up to 30 hops by default, and waits 4 seconds for each reply. The original Unix traceroute by Van Jacobson (1988) sent UDP datagrams to an unlikely port, three per TTL, and stopped when the destination answered “port unreachable”. Both read the same ICMP Time Exceeded messages; pick ICMP or UDP here to match the one you are comparing with.
Does a big jump in time mean that hop is the problem?
Not on its own. From Tokyo the round trip went from 1.24 ms at hop 4 to 75.3 ms at hop 5, and stayed near 75 ms to the destination. That is distance: 74.1 ms of extra round trip allows up to 74.1 ÷ 2 × 299.79 = 11,107 km more travel each way, an undersea cable rather than a fault. A jump that appears at one hop and vanishes at the next is just a slow ICMP reply.
Common Use Cases
Find where a slowdown starts
If the time jumps by 74 ms at hop 5 and stays there to the destination, the delay is on the link into hop 5.
Tell your ISP from the far end
Silent or slow hops 1–3 are near you; problems only past hop 8 sit in the destination’s network.
See how a site routes users
Three probes got three different addresses for wikipedia.org (185.15.59.224, 208.80.154.224, 103.102.166.224): DNS sent each to a different site.
Check a firewall that blocks ICMP
Repeat the trace with TCP to port 443; a web server that drops ping still has to answer a TCP handshake.
CCNA and Network+ study
Watch TTL 1, 2, 3… turn into Time Exceeded replies (ICMP type 11, code 0) and the final Echo Reply (type 0).
Last updated: