Latency vs bandwidth

"The internet is slow" describes two different problems that get fixed in two different ways. Bandwidth is how much data can flow per second — the width of the pipe, measured in Mbps. Latency is how long one round trip takes — the length of the pipe, measured in milliseconds (what is latency). A connection can be enormous in one dimension and terrible in the other.

The pipe and the round trip

Think of a water pipe between two tanks. Widening the pipe moves more liters per second; it does nothing about how long a single drop takes to travel from one end to the other. Satellite links make the distinction vivid: plenty of bandwidth, yet round trips around 600 ms that make every click feel sluggish.

Why more bandwidth doesn't fix latency

Latency is set by distance, the speed of light, and the number of hops a packet crosses — none of which a bandwidth upgrade changes. Moving a client from 100 Mbps to 500 Mbps does not move Frankfurt any closer to Riyadh.

This matters because most interactive work is made of many small, often sequential round trips, not bulk transfer. A web page that performs 30 sequential round trips over a 120 ms path spends 3.6 seconds purely waiting — with almost no data moving. No bandwidth upgrade removes that wait; only shortening the path does: closer servers, fewer hops, caching.

Bandwidth wins in the opposite scenario: backups, video, large file sync. A 500 MB backup finishes about five times sooner on the fatter link, and latency barely matters.

The rule of thumb: bandwidth limits how much, latency limits how fast it feels.

Which one does monitoring measure?

Latency. Monitoring probes are deliberately tiny — an ICMP echo, an HTTP request, a TCP connection attempt — so the recorded time is essentially the round trip, not throughput. UpAll records a response time for every check, charts it on the monitor page, and summarizes it as p95 (percentiles and p95). It does not run speed tests and cannot tell you a link's Mbps.

That's the right trade for a monitoring tool, because latency history with timestamps answers the questions an MSP actually gets asked. When a client calls saying everything is slow, the latency chart shows whether the path or the server genuinely degraded, when it started, and whether it crossed the degraded threshold — or whether the network is fine and the complaint traces back to a bandwidth-hungry transfer or something else entirely. A Ping monitor on the client's router is the cleanest latency record of the line itself.