Online Traceroute

See the path your packets take to any website or IP address, hop by hop, from 8 cities at once: which networks carry them, where the time is added and whether any loss is real. Or trace the path from the world to you.

Updated October 23, 2026

MTR on Globalping probes · networks from Team Cymru · every place checked against the round trip

ExampleMTR · ICMPRecorded Oct 23, 2026 · from 8 cities

wikipedia.org answers from all 8 cities: 1.1 ms from São Paulo, 200 ms from Sydney.

wikipedia.org·6 addresses·in 6 places·5–19 hops·Wikimedia FoundationAS14907·A recorded run: press Trace for yours

wikipedia.org sends the cities to 6 different addresses, in San Jose, Ashburn, Amsterdam, São Paulo, Singapore, and Marseille.
From Johannesburg the time rises 158 ms at hop 5, where Cogent crosses from Johannesburg to London (9,100 km): that's distance, not a problem.
From Sydney, hops 4 to 8 all answer in about 98.1 ms though they're in Sydney, Perth, and Singapore: the packets cross in a tunnel, and each reply goes on to the tunnel's end first.

From 8 cities

0–250 ms · same scale for every city

Johannesburgwikipedia.org

185.15.58.224· 13 hops · 174 ms

1–4RFC 39274 private addresses inside Vultr's networkPrivate address0%0.4 ms0.1–11.1
5206.249.1.25CogentAS174TransitLoss 0%Jitter 53.0159 ms158–159
+158 ms from here: Johannesburg → London, 9,100 km. Distance, not a problem.
6be9736.ccr81.lon01.atlas.cogentco.com154.54.93.42CogentAS174TransitLoss 0%Jitter 53.7160 ms159–160
7ldn-b3-link.ip.twelve99.net62.115.9.28ArelionAS1299TransitLoss 0%Jitter 59.6179 ms179–179
+14.3 ms inside London: the replies come back by a longer way than the packets went.
8ldn-bb2-link.ip.twelve99.net62.115.140.70ArelionAS1299TransitLoss 0%Jitter 61.6185 ms185–185
9prs-bb2-link.ip.twelve99.net62.115.133.239ArelionAS1299TransitLoss 0%Jitter 62.8188 ms188–188
10mei-b5-link.ip.twelve99.net62.115.124.57ArelionAS1299TransitLoss 0%Jitter 62.8188 ms188–188
11wikimedia-ic-370330.ip.twelve99-cust.net62.115.147.133ArelionAS1299TransitLoss 0%Jitter 69.8192 ms188–209
12irb-601.asw1-b12-drmrs.wikimedia.org185.15.58.143Wikimedia FoundationAS14907Destination networkLoss 0%Jitter 61.9182 ms180–184
13wikipedia.org185.15.58.224Wikimedia FoundationAS14907DestinationLoss 0%Jitter 58.0174 ms174–174

Time so farAdded at this hopRouter answers slower0–200 ms, this city's scale

mtr report from Johannesburg

Johannesburg → wikipedia.org
Host                                                                    Loss% Drop Rcv   Avg  StDev  Javg 
 1. AS???   _gateway (169.254.169.254)                                   0.0%    0   5   0.3    0.2   0.2
 2. AS???   100.100.100.1 (100.100.100.1)                                0.0%    0   5   0.3    0.0   0.1
 3. AS???   10.94.0.97 (10.94.0.97)                                      0.0%    0   5   3.0    4.0   3.8
 4. AS???   10.94.0.13 (10.94.0.13)                                      0.0%    0   5   0.4    0.1   0.2
 5. AS174   206.249.1.25 (206.249.1.25)                                  0.0%    0   5 158.6    0.4  53.0
 6. AS174   be9736.ccr81.lon01.atlas.cogentco.com (154.54.93.42)         0.0%    0   5 159.6    0.4  53.7
 7. AS1299  ldn-b3-link.ip.twelve99.net (62.115.9.28)                    0.0%    0   5 178.6    0.1  59.6
 8. AS1299  ldn-bb2-link.ip.twelve99.net (62.115.140.70)                 0.0%    0   5 184.7    0.1  61.6
 9. AS1299  prs-bb2-link.ip.twelve99.net (62.115.133.239)                0.0%    0   5 188.2    0.1  62.8
10. AS1299  mei-b5-link.ip.twelve99.net (62.115.124.57)                  0.0%    0   5 188.2    0.1  62.8
11. AS1299  wikimedia-ic-370330.ip.twelve99-cust.net (62.115.147.133)    0.0%    0   5 192.3    8.4  69.8
12. AS14907 irb-601.asw1-b12-drmrs.wikimedia.org (185.15.58.143)         0.0%    0   5 181.7    1.5  61.9
13. AS14907 wikipedia.org (185.15.58.224)                                0.0%    0   5 173.9    0.1  58.0

How does traceroute work?

Traceroute sends probes that may cross only a set number of routers, one more each time, and every router where a probe runs out replies: that reply gives the router's address and times it, and a reverse DNS lookup finds its name.

The limit is the packet's TTL (time to live; IPv6 calls it the hop limit): each router lowers it by one and drops the packet at zero, sending back an ICMP Time Exceeded message. Probe 1 dies at the first router, probe 2 at the second, until one reaches the destination, which answers itself.

This page runs MTR, which repeats every probe: 5 packets per hop from each city, so each hop gets a best, average and worst time and a loss figure. Every time is a round trip measured by the probe: there and back, plus the router's time to build its reply.

One probe per TTL, San Jose → wikipedia.orgfastest of 5 · Oct 23, 2026

TTL 1Oracle Cloud0.2 ms
TTL 2Equinix0.4 ms
TTL 3Equinix1.8 ms
TTL 4Wikimedia Foundation2.0 ms
TTL 5wikipedia.org1.7 ms

The probe, its TTL counting down at each router
TTL reaches 0: the router drops it
The router's reply: ICMP Time Exceeded

Each line is one probe. The first expires at the first router, the next at the second, and so on: every router that drops a probe says so, and the time of its reply is the hop's time. A real MTR run from San Jose (Globalping, Oct 23, 2026): 5 hops, 1.7 ms to wikipedia.org.

Online traceroute result: the path to wikipedia.org from 8 cities on a world map, with each hop's network and time — A traceroute to wikipedia.org from 8 cities, hop by hop. Select a number to see what each part shows.

How do you read a traceroute?

Read it top to bottom: each line is one router on the way, with how long its reply took. What matters is where the time goes up and stays up, and whether the destination answers.

A high time on one line that the next lines don't share is a router that is slow to reply, not slow to forward: building the reply is low-priority work for its processor. Loss that disappears after a hop is a router limiting its replies. Only time or loss that carries on to the last line affects what you feel. The address on each line is the interface the probe arrived on, so the same router can show a different address on another path.

Mumbai → wikipedia.org, read line by lineMTR, 5 packets a hop

#HopNetworkLossAvg

11140.91.204.30Oracle Cloud0%0.6

2125.17.12.178Airtel20%17.5

3125.17.12.177Airtel0%2.3

4116.119.42.61Airtel0%355.3

5* * *––

6ethernet-1-55-100.asw1-603-eqsin.wikimedia.org4Wikimedia Foundation0%57.4

7wikipedia.orgWikimedia Foundation0%562.6

1Hop numberThe TTL: the first router is 1.
2The networkWho runs the router: its AS number and brand.
3The jumpWhere the time rises and stays up: usually distance.
4The router's nameIts PTR name; codes in it often name the city.
5The destinationThe last line: the time a visitor's packets really take.

A real run (Globalping, Oct 23, 2026). The time that counts is the one that stays: a hop slower than the hops after it only answered slowly.

Why does traceroute show asterisks (* * *)?

An asterisk is a probe that got no reply in time. Usually that router just doesn't answer probes; it only matters when the stars last until the end.

Each tool waits a few seconds per probe (5 s on Linux and macOS, 4 s in tracert) and prints * when nothing comes back. Routers may be set never to answer, or to limit how many replies they send, and firewalls often drop the probes or the replies. A middle hop of stars with answers after it means the packets passed.

A router that never answers

Normal

Hops 3–6 stay silent, but hop 7 and the destination answer: those routers pass packets on and just don't reply.

São Paulo → wikipedia.org · hops 2–8

A router that answers some probes

Normal

Hop 7 replied to 2 of 5 probes, and the hops after it to all of them: it limits its replies; nothing is lost.

São Paulo → wikipedia.org · hops 5–9

The far end drops the probes

Try TCP 443

After hop 8 nothing answers, up to the 20-hop limit: the destination's network or a firewall drops these probes.

Our server in Mumbai → bbc.co.uk (UDP) · hops 5–20

Three real traces. An asterisk is a probe that got no reply within the wait: only a stop that lasts to the end is about the destination.

Traceroute vs ping vs MTR vs pathping

Ping times one destination; traceroute shows every router on the way; MTR keeps tracing so each hop gets real loss and jitter figures; pathping is Windows' slower version of the same idea.

Classic traceroute sends 3 probes per hop and moves on, so a hop that drops one probe looks like 33% loss. MTR repeats every hop round after round, and pathping sends 100 queries per hop, 250 ms apart, before it reports. To test a whole connection rather than one route, use the ping test or the packet loss test.

Shows the path	pingping host	traceroutetracert host · traceroute host	MTRmtr host	pathpingpathping host
Shows the path		every hop	every hop, live	every hop
Time per hop		3 probes a hop	average, best, worst	after 100 queries
Loss per hop		a few probes can't tell	every round counts	100 probes a hop
End-to-end time and loss	yes	the last line	the last line	yes
How long it takes	4 s	5–30 s	runs until stopped	about 25 s a hop
Use it when	you want one number	you want the route	you suspect loss	you're on Windows and need loss

What each tool measures. MTR is traceroute that keeps going, so its per-hop loss means something; this page runs MTR from every city. pathping's time is arithmetic from its defaults: 100 queries 250 ms apart per hop.

What port does traceroute use?

None of its own: Linux and macOS traceroute send UDP to port 33434 and up, Windows tracert sends ICMP echo requests, and TCP mode sends SYN packets to a port you choose (443 gets through most firewalls).

The UDP probes aim at ports nobody listens on, so the destination answers "port unreachable" and the trace ends there. When a trace stops before a website that loads fine, switch to TCP 443 here: firewalls let web traffic through more often than UDP or ICMP. To check a port itself, use the Port Checker.

What each tool puts in the probedefault first

Linux traceroutetraceroute hostIP · TTL 1, 2, 3…UDPdst port 33434 and up
macOS traceroutetraceroute hostIP · TTL 1, 2, 3…UDPdst port 33434 and up
Windows tracerttracert hostIP · TTL 1, 2, 3…ICMP Echo Requestno port
ICMP modetraceroute -I · mtrIP · TTL 1, 2, 3…ICMP Echo Requestno port
TCP modetraceroute -T -p 443 · mtr -T -P 443IP · TTL 1, 2, 3…TCP SYNdst port 443
This pageGlobalping MTRIP · TTL 1, 2, 3…ICMP · TCP 443 · UDPyou choose

Traceroute has no port of its own: Linux and macOS send UDP to ports from 33434 up (IANA registers only 33434, for traceroute use), Windows sends ICMP echo requests. Firewalls that drop those often let TCP to port 443 through, because it looks like a visit to a website.

How to run the traceroute command on Windows, Mac and Linux (tracert)

On Windows, open Command Prompt (cmd) and type the tracert command followed by the address; on macOS and Linux, open Terminal and type traceroute. Add -d (Windows) or -n to skip name lookups and finish faster.

tracert sends ICMP and waits up to 4 s per probe; macOS and Linux send UDP. mtr (on macOS through Homebrew, on Linux from the package manager) gives the per-hop loss this page shows. A trace from your own computer shows the first hops inside your network, which an online trace can't see; this page shows how the rest of the world reaches the same host.

Windows
macOS
Linux
MTR

C:\> tracert -d example.com

Three probes per hop, 30 hops at most; -d skips name lookups and finishes faster.

Our server in Mumbai → bbc.co.ukUDP · 1 probe a hop · Oct 23, 2026

$ traceroute -m 20 -w 2 -q 1 bbc.co.uk
 1  103.118.16.254 (103.118.16.254)  3.815 ms
 2  195.250.22.6 (195.250.22.6)  0.293 ms
 3  delhi-203.197.239-201.vsnl.net.in (203.197.239.201)  0.880 ms
 4  49.44.18.38 (49.44.18.38)  1.335 ms
 5  10.70.14.97 (10.70.14.97)  2.130 ms
 6  *
 7  192.168.65.138 (192.168.65.138)  16.185 ms
 8  49.44.53.3 (49.44.53.3)  17.261 ms
 9  *
10  *
11  *
12  *
13  *
14  *
15  *
16  *
17  *
18  *
19  *
20  *

1Hop 1 is the server's own gateway, the first router on the way.
2A single * is a router that didn't reply; the trace carries on to the next hop.
3No answer after hop 8, up to the 20-hop limit: the far end drops these UDP probes.

Our server's recorded run to bbc.co.uk (Oct 23, 2026, UDP, one probe a hop), printed the way traceroute prints it. The commands are the same in every language.

Why does latency jump at one hop?

Usually because that hop is where the path crosses a long distance, often an ocean or a continent. Light in fibre adds about 1 ms of round trip for every 100 km, so a jump that matches the distance is normal.

Cables don't run in straight lines, so allow about 12.5 ms per 1,000 km of map distance. A jump more than about twice what the distance needs, or one that appears without the route moving, points at a detour, a busy link or a longer way back: each reply can return by a different route from the one the probes took. Inside an MPLS tunnel every hop can show the tunnel's end time, so several hops in different cities report the same number.

Where the time is added: Johannesburg → wikipedia.orgslow replies set aside

Is it distance? 1 ms of round trip per 100 km1 jump, measured on Oct 23, 2026

Johannesburg → London9,100 km · at least 91 ms+158distance

The least the distance allows (light in fibre)
What the jump measured

Light in fibre covers about 200 km a millisecond, so a round trip costs at least 1 ms per 100 km of straight-line distance; cables are longer than the straight line. A jump within about twice that floor is distance. Each step is the lowest time at that hop or after it, so a router that is only slow to reply doesn't make a step.

Traceroute stops at a hop or says request timed out

A trace that stops usually means the destination, or a firewall in front of it, ignores traceroute's probes: if the site loads, the network is fine. Try TCP 443 before blaming the route.

Stars from one hop to the end, but the site loadsThe most common case

The destination or a firewall in front of it drops traceroute's probes. Run it again over TCP 443: it looks like a visit to the website, so it usually gets through. A site that loads is reachable; only the probes went unanswered.

Request timed out on every hopFrom hop 1 onwards

Your own firewall, VPN or network blocks the probes or the replies. Try another protocol, or run the trace from this page, outside your network.

It stops right after hop 1 or 2Your router answers, nothing after it

The connection to your provider is down, or its network drops probes. If websites don't load either, restart the router and call your provider with the trace.

Stars in the middle, later hops answerSilent routers

Those routers don't reply to probes, or limit their replies. The packets passed through them: nothing is wrong.

The same two addresses repeat to the endA routing loop

Two routers send the packets back and forth until the TTL runs out. It's a fault in a network on the way: send the trace to the provider whose network it's in.

!H, !N or !X after a timeThe router says it can't reach it

Host unreachable, network unreachable or administratively prohibited: a router answered that it won't or can't forward the packets. The address may be wrong, or blocked on purpose.

Find what you see and open it for the reason and the next step. The dots are each hop: filled answered, hollow got no reply.

How do I trace the route to my own IP address?

Choose Trace to me and press the button: probes in several cities trace back to your public IP address, so you see how the internet reaches your provider and where the time is added on the way to you.

The last hops before your address usually stay silent: home routers don't answer probes. Globalping publishes every measurement, so your IP address becomes part of a public result: the trace starts only when you press the button, and we don't keep your address.

3 places trace to one home line (Jio, India)MTR · 5 packets a hop

MumbaiOracle Cloudstops at hop 2
FrankfurtOracle Cloud211 ms
AshburnOracle Cloud245 ms

YouJio · AS55836 · the last hops before you stay silent, as home lines usually do203.0.113.7

The path to you is a different path from the one your own traceroute takes outwards: each direction is routed separately. A recorded run to a home line; the address is a documentation one.

Online traceroute: questions and answers

What is traceroute used for?

Traceroute shows the route packets take to a website or IP address and how long each router on the way takes to answer. It's used to find where a connection slows down or stops: inside your provider, on a long-distance link, or at the destination. This online traceroute runs it from 8 cities at once, so you can compare routes.

How do I run a traceroute or tracert online?

Type a website or IP address above and press Trace. Probes in San Jose, Ashburn, London, São Paulo, Mumbai, Tokyo, Sydney and Johannesburg run MTR to it at the same time, with 5 packets per hop. Each city's path appears hop by hop with its networks, times and loss, and on a world map.

What does tracert do, and is it the same as traceroute?

Yes in purpose, different in detail. tracert is the Windows command: it sends ICMP echo requests, three per hop, up to 30 hops. traceroute on Linux and macOS sends UDP probes by default. Both raise the TTL one step at a time and list the routers that answer. This page can send ICMP, TCP or UDP probes.

What does * * * mean in a traceroute?

Each asterisk is a probe that got no reply within the wait time. Many routers don't answer traceroute probes or limit their replies, so stars in the middle of a trace are normal when later hops answer. Stars that last until the end mean the destination or a firewall drops the probes; try TCP 443 here.

Why does my traceroute stop before the website?

Usually the website's network or a firewall drops traceroute's UDP or ICMP probes, while ordinary visits still get through. If the site loads, the network is fine. Run the trace again with the protocol set to TCP 443: it looks like a web visit, so it reaches the destination more often.

Is high latency at one hop a problem?

Only if it carries on to the destination. A router that's slow to answer probes shows a high time on its own line, while the hops after it are faster: that's cosmetic. A jump that stays to the end is usually distance, about 1 ms of round trip per 100 km of fibre; this page checks each jump against it.

Can I traceroute to my own IP address?

Yes: choose Trace to me. After you press the button, probes in several cities trace back to your public IP address, showing how the internet reaches your provider. The last hops usually stay silent because home routers don't answer probes. Globalping publishes every result, so your address becomes part of a public measurement.

What is MTR, and how is it different from traceroute?

Traceroute sends a few probes per hop once and stops. MTR keeps sending them round after round, so each hop gets loss, jitter and best, average and worst times from many packets. That makes per-hop loss meaningful. This page runs MTR with 5 packets per hop from every city.

How many hops is normal for a traceroute?

There's no standard number: it depends on how many networks and routers sit between the probe and the destination. Nearby sites can be 4 to 6 hops away, distant ones 15 or more. Tools stop at a maximum: 30 hops in tracert, Linux traceroute and MTR, 64 in macOS traceroute.

Is this online traceroute free, and what is stored?

It's free, with no sign-up. The traces run on Globalping, a free probe network by jsDelivr: each city uses one of your network's 250 free tests an hour (dnsrobot.net's own when your browser can't reach Globalping), and Globalping keeps the results public. We look up each hop's network and place for the page and keep nothing about you.

Missing something?Suggest a feature or a new tool

Online Traceroute