What Ping Time Actually Measures

Ping is the round-trip time for a small packet to reach a remote host and come back. The classic ping utility sends an ICMP echo request and waits for the echo reply, reporting the elapsed time in milliseconds. That single number bundles together four physically distinct delays, and separating them is the difference between diagnosing a problem and guessing at one.

Propagation delay is the time the signal spends in transit. It is set by distance and by the speed of the signal in its medium, and no amount of money can reduce it below the speed of light. Serialization delay is the time to clock the packet's bits onto the wire, which is packet size divided by link rate. Processing delay is what each router spends parsing headers and making a forwarding decision. Queuing delay is time spent waiting behind other packets in a buffer, and it is the only one of the four that varies wildly from moment to moment. Propagation and serialization are computable from first principles. Processing is a small, fairly stable per-hop constant. Queuing is the wildcard, and it is where nearly all troubleshooting effort belongs.

Why Distance Sets a Floor You Cannot Argue With

Light in a vacuum covers 299,792 kilometres per second, or about 300 km per millisecond. Inside a single-mode fibre the glass core has a refractive index near 1.47, so light slows to roughly 204 km per millisecond. That figure is the single most useful number in network latency work. It means a signal needs about 4.9 milliseconds per 1,000 kilometres of fibre one way, or 9.8 milliseconds per 1,000 kilometres round trip.

London to New York is 5,570 km in a straight line. At fibre speed, that alone is 27 ms one way and 55 ms round trip before a single router is involved — and real transatlantic cable routes are longer than the great circle, so measured values sit closer to 70 ms. This is why content delivery networks exist. No protocol optimisation, hardware upgrade, or premium ISP plan can beat physics; the only way to reduce propagation delay is to move the content closer to the user. A CDN edge node 50 km away delivers a 1 ms propagation delay that a perfectly tuned connection to a server 5,000 km away can never approach.

Route Circuitousness: The 40% You Forget

Fibre does not travel in straight lines. It follows railway rights of way, motorway easements, existing conduit, and seabed routes chosen to avoid trawler damage and geological hazards. The ratio between actual cable length and straight-line distance is called circuitousness, and on real backbone routes it typically measures between 1.2 and 1.6. Using 1.4 as a default is a reasonable planning assumption for continental paths.

This factor matters more than most people expect. A 1.4× route penalty on a 5,000 km path adds 2,000 km of glass, which is nearly 20 ms of round-trip delay. It is also why certain city pairs have surprisingly good or bad latency relative to their map distance: two cities on a major submarine cable landing route can beat two closer cities whose traffic has to detour through a distant exchange point. Peering geography, not physical geography, often determines the path.

Reading Your Own Ping Against This Estimate

Use this calculator to establish what your path should cost, then compare it to what ping or your game client reports. If the two agree within 10 or 20 percent, your connection is behaving normally and the latency you see is structural — the only fix is a closer server. If your measured ping runs far above the estimate, the surplus is diagnosable.

Ping that is fine when idle but spikes when anyone uploads is bufferbloat: an oversized queue on your uplink filling with data. The fix is a modern queue discipline such as fq_codel or CAKE on your router, not a faster plan. Ping that is consistently high and jittery on Wi-Fi but clean on Ethernet points at radio retransmissions, channel congestion, or distance from the access point. Ping that is stable but far above the estimate usually means the route is not taking the path you assumed — traceroute or mtr will show you where it actually goes, and the answer is often a detour through an exchange in another country. Finally, remember that ICMP is sometimes rate-limited or deprioritised by routers, so a high ping to a device that forwards traffic perfectly well is not always a real problem.