An example of throughput vs. latency
In software engineering jargon, people often speak of “performance” without defining specifically what they mean. Improving performance usually involves “increasing throughput”, or “reducing latencies”.
Throughput is measured in “units over time”, for example, “bytes per second”, or “cars passing by every day”.
Latency, meaning “how long something takes”, is measured in “time per operation”. For example, how long it takes to load a website or send an image.
I remember my early days of trying to understand the difference. My breakthrough came when somebody gave the following example:
“Suppose you want to transfer 10 petabytes of hard drives from Europe to America. For this example, that would be ~10k hard drives.
One approach would be to transfer one hard drive at a time across the Atlantic subsea cables. The latency of sending each hard drive would be fairly low (maybe an hour), but throughput very low (number of hard drives uploaded per hour).
