RSTP (Rapid Spanning Tree Protocol)
What it fixes
STP prevents loops but is too slow: recovering from a link failure takes 30-50 seconds (15 listening plus 15 learning).
★★★ Traffic is down for those 30 seconds. For voice, storage or databases, 30 seconds is an outage.
RSTP brings it down to 1-3 seconds, using exactly the same frame format as STP (only the version number is 2).
Where the speed comes from
★★★ (1) Handshake instead of waiting: rather than sitting out timers, a switch agrees with its neighbour directly using the proposal / agreement bits in the BPDU.
★★★ (2) Edge ports forward immediately: a port facing an endpoint does not wait on the spanning tree at all.
★★★ (3) A pre-designated backup: the new Alternate role knows in advance that it is the backup, so on failure it takes over without recomputing the whole tree.
★ Common misconceptions
★★★ "RSTP is a different protocol; you must upgrade everything" -- it interoperates with STP (falling back to STP on those segments). But one legacy STP switch drags that segment back to STP speed.
★★ "Enabling RSTP makes it fast" -- only if edge ports are configured as edge. Otherwise host-facing ports still wait, which is the usual reason for "RSTP is still slow" in the field.
★★ "Fast means loops no longer matter" -- RSTP only converges faster; it still blocks ports and does nothing about the idle half of your bandwidth.