High Latency Troubleshooting: Find Where the Delay Starts
Separate propagation delay, congestion, routing changes and endpoint delay without guessing.
Start with a normal baseline
Latency only has meaning in context. A 35 ms round trip can be excellent across a regional backbone and poor inside a campus. Compare current measurements with historical data or a known-good peer path.
Stable extra delay often points toward a longer route or path change; variable spikes more often suggest queuing, loss, CPU pressure, or a busy access segment.
Find the first point where added delay persists
Trace the path and look for the first hop where the latency increase remains visible in subsequent hops. A single slow router reply is not enough because control-plane responses can be deprioritized.
Check the reverse direction when possible. Asymmetric routing may put each direction on a different provider, firewall, or geographic path.
Separate network delay from application delay
Inspect utilization, drops, shaping, policing, and recent routing changes near the first persistent delay. A circuit can be below 100 percent average utilization while short bursts create latency.
Compare network round-trip timing with application response timing. If TCP reachability is quick but the application is slow, shift attention to the server, database, proxy, TLS negotiation, or other application dependencies.