Tracking area — The granularity at which the core knows where you are

Inside its TAI list, the device says nothing. It may reselect freely, cell to cell, all day, and the core is never told.

Crossing out of it, the device must register — a Mobility Registration Update — and is given a new list.

This is one trade, and the size of the TAI list is the dial.

Make the list large — the device reports rarely, and signalling is cheap. But every page must be broadcast across a large area, so paging load grows with the number of cells and with how often devices are paged.

Make the list small — pages are cheap because they go to few cells. But devices cross boundaries often, and each crossing is a registration: a radio connection, a NAS exchange, core signalling. Registration load grows.

Both costs are paid in the same currency and the optimum depends on how much devices move against how often they are reached. It is the same trade as periodic registration makes on the time axis, made here on the space axis.

The list is per-device, and that is the clever part. It is not a fixed partition of the map into zones: the AMF gives this device a set of areas chosen for it. A device that repeatedly crosses the same boundary can be given a list containing both sides, and the ping-ponging registrations stop — without changing anything for anyone else. Two devices standing next to each other may hold different lists.

Before this concept, the hierarchy says to learn the following — the full chain, in order: