How much disk does a year of traffic take?

The question comes up on every migration, usually from whoever manages the backups. It has a satisfying answer.

A tool that writes one row per pageview grows in step with your traffic. Ten times the visitors, ten times the table, and eventually a nightly backup that takes long enough to notice.

A tool that writes one row per hour per thing it counts does not. The row for a given page, on a given day, in a given hour, gets updated rather than duplicated. A site with a hundred thousand views a day and one with a hundred use roughly the same disk, because they have roughly the same number of pages.

On this site a full year of history is a few tens of megabytes, and most of that is the list of distinct paths rather than the counts. The practical consequence is that nobody has to choose between keeping history and keeping backups quick.