Online traceroute and MTR from 22 locations.

Trace the route to any server or website from 22 monitoring nodes across 13 countries. Every hop comes back with its packet loss, latency and network, so you can see where on the path a slowdown or a drop begins.

22 Monitoring nodes
13 Countries
$0 No account

Trace the route to any public hostname or IP address from 22 locations at once. Choose ICMP or TCP and how many cycles to run; each location reports every hop with its loss and latency.

Protocol
Try:

Note: All checks are public. Create a free account to keep your checks private and get an email if your site stops responding.

What is Traceroute?

Traceroute shows the path packets take from one machine to another across the internet. Every packet carries a Time to Live (TTL) counter that each router on the way lowers by one. A router that sees it reach zero drops the packet and sends back an ICMP Time Exceeded message. Traceroute sends probes with a TTL of 1, then 2, then 3 and so on, so each router along the route reveals itself in turn, along with how long its reply took.

The result is a numbered list of hops from the source to the destination. It answers questions a ping cannot: which networks the traffic crosses, where the route leaves one provider for another, and at which hop the latency jumps or the replies stop.

Emercom runs the trace from 22 monitoring nodes in 13 countries, so you see the route to your host from many networks instead of only your own.

A ping tells you whether the destination answers and how long the round trip takes. A trace tells you which way the packets went to get there, and where on the way the delay or the loss appears.

MTR: Traceroute + Ping

MTR (originally Matt's traceroute, now My Traceroute) runs a traceroute and then keeps probing every hop for a number of cycles, collecting statistics as it goes. Instead of one reply per hop you get a sample: the share of probes each hop lost, and its last, average, best and worst round-trip times. That is what makes MTR useful for problems that come and go, which a single traceroute pass easily misses.

Every Emercom location runs mtr --report with AS lookup turned on, so each hop also carries the autonomous system (AS) number of the network it belongs to. You choose the probe protocol (ICMP, or TCP to a port you pick, 443 by default) and the number of cycles: 5, 10 or 20, with 10 as the default.

Alongside the trace, the same location runs a ping and a DNS lookup, so you can see what the name resolved to from there and how the destination answers a plain ping.

Reading MTR Results

Each location returns its own report: one row per hop, numbered from that location outward to your host. These are the columns:

Column Name What it means
Host Hop The router that answered at this hop: its name and address, with the AS number of its network.
Loss% Packet Loss Share of probes to this hop that got no reply.
Snt Sent Probes sent to this hop, one per cycle.
Last Last RTT Round-trip time of the most recent reply, in milliseconds.
Avg Average RTT Mean round-trip time across all replies from this hop.
Best Best RTT Lowest round-trip time seen.
Wrst Worst RTT Highest round-trip time seen.
StDev Standard Deviation How much the round-trip times vary from probe to probe. A high value means unstable latency at that hop.

Read Loss From the End

Start at the last row. If the destination shows 0% loss and steady latency, the path works, whatever the hops in the middle show. If the destination shows loss, walk back up the list and find the first hop where the loss begins and then carries on to the end: the problem sits at that hop or on the link just before it.

Latency Jumps

Latency rises as the path gets longer. A jump that stays on every later hop marks a slow or long link, such as an ocean crossing or a congested handover between two networks. A jump on a single hop that the later hops do not repeat is that router replying slowly to probes, not a slow path.

Hops That Do Not Answer

A row of ??? is a hop that never replied. Many routers are set up not to answer probes at all, so one silent hop in the middle of a working path is normal.

Rows Past the End of the Route

When the destination never answers, mtr keeps probing up to its hop limit, and those probes can circle back through routers that already appeared in the list. Emercom dims every row past the point where the real route ends, so a repeated router or a run of silent hops at the bottom is not mistaken for part of the path.

Hover a Hop for Its Network

Hover a hop in the result to see the network that owns its address and the approximate location of that address. Private and reserved addresses, common in the first hops inside a provider, are labelled as such.

Why Trace Routes From Multiple Locations?

A trace from your own computer shows one path: the one your provider picks. Visitors in other countries reach your server over different networks, and a problem on one of those paths stays invisible from yours. Tracing from many locations at once shows you:

  • Regional routing problems. A broken or congested path from one region shows up as loss or high latency from the locations there, while the others look normal.
  • Where the paths meet. Traces from different places converge as they approach your server. Loss that starts after the point where they meet affects every visitor; loss before it affects only some.
  • Which networks carry the traffic. The AS numbers on each hop show which providers sit between a region and your server, and where the route hands over from one to the next.
  • Anycast & CDN routing. For an anycast address or a CDN, each location should end at a nearby edge. A trace that crosses an ocean to get there shows users being sent the long way.
  • A baseline to compare against. A trace taken while everything works records the normal path and latency, so a trace during an incident shows what changed.

Emercom runs the trace from its locations on 5 continents and returns every report on one page, under a link you can share.

ICMP or TCP?

The protocol decides what kind of packet MTR sends as a probe. The route is usually the same either way; what changes is whether firewalls on the path, and the destination itself, let the probes through and answer them.

ICMP

The default. Probes are ICMP echo requests, the same packets ping sends. Use it for a general look at the path and for hosts that answer ping.

TCP

Probes are TCP connection attempts to the port you choose, 443 by default. Use it when ICMP is filtered on the path or at the destination, or to see how packets that look like your real traffic are treated on the way to the service's port.

If a trace stops short over ICMP, run it again over TCP to the port your service listens on. If TCP reaches the destination, the earlier stop was ICMP filtering, not a broken path.

Common Uses for MTR

Most traces start with a complaint that something is slow or dropping. These are the situations where a multi-location trace answers the question fastest.

Packet Loss Reports

When users report drops or timeouts, a trace from their region shows whether loss appears on the way and at which hop. MTR output is also what network and hosting providers usually ask for when you open a ticket.

Slow Regions

When a site is fast in one country and slow in another, traces from both show where on the path the extra latency comes in.

Provider & Routing Changes

After switching providers, adding an upstream or changing BGP announcements, traces from many locations confirm the traffic takes the routes you expect.

Firewall & Filtering Checks

Compare an ICMP trace with a TCP trace to the same host to tell a filtered protocol apart from a path that really breaks.

Traceroute Myths That Won't Die

Trace output looks precise, which makes it easy to over-read. These are the misreadings that send people chasing the wrong router.

"Loss at a hop means that router drops my traffic."

Reality: Routers often limit how fast they answer probes aimed at themselves while forwarding everything else normally. Loss that does not carry on to the following hops and the destination is that limit at work, not lost traffic.

"Latency always goes up hop by hop."

Reality: A router can be slow to answer a probe and still forward traffic quickly, so a middle hop can show more latency than the hops after it. Only an increase that stays on every later hop reflects the path.

"A silent hop means the route is broken."

Reality: Plenty of routers never answer trace probes. If the hops after the silent one reply, traffic is passing through it. Only silence that runs to the end of the trace, with no reply from the destination, says the probes stopped getting through.

"A trace shows the way back too."

Reality: It shows only the forward path, from the probing location to the target. Replies can return over a different route, and their travel time is part of every number in the report. To see the return path, run a trace from the other end.

"If ping is blocked, the host is down."

Reality: Many hosts and firewalls drop ICMP and still serve traffic. A TCP trace to the service's port often reaches a destination that ignores ping.

"Every device on the path shows up as a hop."

Reality: Only devices that lower the TTL and answer appear. Switches never show up, and MPLS tunnels or some firewalls can hide several routers behind a single hop.

Frequently Asked Questions

Quick answers to the most common questions about traceroute, MTR and this tool.

Traceroute makes a single pass and lists the routers that answer at each hop. MTR keeps sending probes for several cycles and keeps statistics for every hop: the share of probes lost and the last, average, best and worst round-trip time. That makes MTR better at showing loss and latency that come and go, which a single traceroute pass can miss.

Most routers forward traffic on a fast path but answer probes addressed to themselves on a slower one, and many limit how often they reply. A hop can then show loss while every packet passing through it arrives. If the hops after it and the destination show no loss, the path is fine. Loss that starts at one hop and continues all the way to the destination is the kind worth investigating.

Use TCP when the destination, or a firewall on the way, drops ICMP. In TCP mode the probes are connection attempts to the port you choose (443 by default), the same kind of packet a web or API client sends first, so they often get through where ICMP is filtered. Choose the port your service actually listens on.

MTR marks a hop that never replied with ???. A run of them at the end of the trace, with no reply from the destination, means the probes stopped getting answers: the destination may be down, a firewall may drop the probes, or the host may not answer that protocol. Try the other protocol, and compare locations: if only some of them fail, the problem is on their paths rather than at the destination.

Each cycle sends one probe to every hop, and cycles run about a second apart. 10 cycles, the default, suits most traces. Choose 5 for a quicker answer, or 20 to steady the averages and catch loss that comes and goes. With 10 cycles, every lost probe adds 10% loss at that hop.

Yes, on a monitor. When a check of an HTTP, port or ICMP monitor fails, the location that saw the failure runs MTR over both ICMP and TCP, together with ping and a DNS lookup, and saves the output as a Forensic Snapshot. The MTR tool on this page is a one-time check.

Capture a trace the moment a check fails

The MTR tool runs once, when you ask. Set up an HTTP, port or ICMP monitor instead, and when a check fails, the location that saw the failure runs MTR over ICMP and TCP, with ping and a DNS lookup, and saves the output as a Forensic Snapshot. Alerts go out by email, SMS, voice call or Telegram.

Choose what to allow. Essential cookies keep you signed in and protect forms; they are always on.