Start with ICMP
Use the standard preset for a first look. If it gets no replies, try TCP 443.
THE PATH, MADE VISIBLE
Every hop tells a story.
Trace the route. Find the friction. Understand what happens in between.
Route explorer
READYA little clarity, one hop at a time.
Enter a public hostname or IP address to see the route from sw3.
DESTINATION REPLY LOSS—Waiting for a destination reply
DESTINATION AVG. RTT—Round trip, including the return route
OBSERVED HOPS—Starting…
Discovering route…
| HOP | ROUTER / NETWORK | LOSS | SENT | RECV / WAIT | LAST | AVG | BEST | WORST | STDEV | GEOMEAN | JITTER | JIT. AVG | JIT. MAX | JIT. EWMA | RTT HISTORY |
|---|
Latency values are in milliseconds. A silent hop can still forward traffic. Hop RTTs include different return paths.
Equivalent command & measurement notes
One MTR process produces each result. Statistics are calculated from its sent/reply events. Jitter is the absolute difference between successive received RTTs; mean jitter includes the initial zero sample. Standard deviation uses n−1. Interarrival jitter uses an EWMA with a 1/16 gain. These calculations retain floating-point precision, so tiny differences from MTR's integer-rounded display are expected. Loss excludes probes still inside the reply window; a completed run settles all unanswered probes. A stopped or failed run retains pending probes.
Hop statistics combine the responding addresses at that TTL. Expanded rows retain observed alternate addresses and MPLS labels; this is not a complete map of every possible ECMP path.
Read beyond one red number
Intermediate routers may limit replies. Compare later hops and the destination before drawing conclusions.
Compare like with like
Keep protocol, IP version, source, and timing the same when comparing two tests.
Queries are sent only when you press Look up, directly from your browser to the selected resolver. Answers reflect that resolver's view and cache.
Inspect the NS records visible to public recursive resolvers and resolve each nameserver's address. This is a delegation check, not a DNSTT connection test or an authoritative DNS trace.
For delegated services
A common layout is an NS record for t.example.com pointing to ns.example.com, with A/AAAA records for that nameserver. A nameserver inside the delegated zone needs suitable parent glue records.
A published AAAA record should point to a working IPv6 service. DNS records alone cannot confirm UDP/53 reachability or tunnel authentication.
A little more signal.
Packet Lab is a small, self-hosted networking workbench. Subnet calculations, CIDR operations, and endpoint parsing run in your browser. No analytics, third-party fonts, or remote scripts are loaded.
MTR destinations and options are sent to this server to run the test. Up to 32 recent tests are kept in memory for at most 15 minutes from their start, subject to service restarts. Reverse DNS and ASN lookups are optional. DNS Desk and Delegation Inspector contact your selected public resolver when requested.
Your IP address is used for service limits. The supplied Nginx configuration disables site access logs; system and network infrastructure may have their own logs. Trace tokens stay in page memory. Appearance is the only preference saved in local storage.
MTR measures replies to probes from this server. It does not measure your browser's route, application throughput, DNS tunnel performance, or every forwarding path. Routes and response policies can vary by protocol and over time.