Refreshing a network without stopping the organisation

A network refresh is rarely difficult technically. It goes wrong on sequencing, on parts nobody costed, and on the discovery that the new switch will not take the old optics. All three are avoidable at the quotation stage.

Count the optics before you count the switches. Transceivers are frequently a large share of the cost and the longest lead item, and compatibility is not a given between generations or vendors. Get the optic part numbers on the same quotation as the switches so the whole thing arrives together.

Check the PoE budget against what is actually plugged in. Access points, cameras, phones and door controllers all draw from the same budget, and the newer generation of each usually draws more than the one it replaces. A switch that is fine on port count can be short on watts.

Order the boring parts at the same time. Patch leads in the right lengths and colours, rack screws, blanking plates, cable management, PDUs and the correct power cords for the region. These are cheap, they are always forgotten, and a missing power cord stops a cutover just as effectively as a missing switch.

Sequence by blast radius, not by convenience. Start with the least critical closet, prove the configuration, then move up. Keep the old equipment on site and configured until the new one has survived a full working week.

Plan the disposal at the start. Decommissioned equipment has to be wiped, recorded and removed. Institutions with asset registers need it recorded properly, and organisations with data-bearing devices need a certificate. Deciding this at the end is how equipment ends up stacked in a corridor for six months.