← Module 02 · Layer 2: Data Link

RSTP & MST

Same job as classic STP (Spanning Tree Protocol): one active path per segment, loops prevented, but reaching that state in under a second instead of the better part of a minute.

RSTP (802.1w): The same tree, a faster handshake

802.1D blocks a port and then just waits, stepping through timers because it has no way to know for certain the topology is actually safe yet. RSTP (Rapid Spanning Tree Protocol) replaces that waiting with an active conversation between neighboring switches: a proposal/agreement handshake, where a switch proposes "I'll be your designated port," the neighbor checks that its own ports are safe to synchronize, and agrees. That back-and-forth typically finishes in well under a second, instead of 30–50.

SW1 SW2 PROPOSAL: "I'll be designated here" AGREEMENT: "confirmed, forwarding immediately"
RSTP negotiates instead of waiting out a timer. A port moves to forwarding as soon as its neighbor confirms it's safe.

RSTP also adds two port roles beyond 802.1D's blocking/root/designated set, so a failover has less work to do:

Here's what those roles actually look like once they're assigned on a real three-switch topology, with one redundant link between SW2 and SW3:

ROOT BRIDGE SW1 SW2 SW3 PC DP RP DP RP DP AP Edge RP = root port · DP = designated port · AP = alternate port (discarding, no traffic) Edge skips straight to forwarding, no negotiation needed
SW1 is root, so both its ports are designated. SW2 and SW3 each use their lowest-cost port to SW1 as their root port. On the SW2–SW3 link, SW2's port wins as designated and SW3's port becomes alternate: physically connected, but discarding, exactly like blocking in classic STP.
STP (802.1D)RSTP (802.1w)
How it convergesPassive, waits out fixed timersActive, proposal/agreement handshake
Typical convergence30–50 secondsUnder a second
Port statesBlocking, Listening, Learning, ForwardingDiscarding, Learning, Forwarding
Extra rolesNoneAlternate, Backup, Edge

MST (802.1s): One tree isn't always the goal

Running a single spanning tree across every VLAN (Virtual Local Area Network) means every VLAN blocks the exact same link, which wastes the bandwidth of every other physical link for every VLAN that isn't using it. Cisco's PVST+ (Per-VLAN Spanning Tree Plus) solves that by running one full spanning tree instance per VLAN, so different VLANs can block different links and actually load-balance. But that means hundreds of VLANs can mean hundreds of spanning tree instances, all consuming CPU (Central Processing Unit) cycles and BPDUs (Bridge Protocol Data Units).

Multiple Spanning Tree (MST) is the middle ground: VLANs are grouped into a small number of instances, and each instance runs its own independent tree. You get the load-balancing benefit of per-VLAN trees, the CPU and BPDU overhead of only a handful of instances instead of hundreds.

WHAT TO ACTUALLY REMEMBER

In real deployments, RSTP (or a vendor's rapid variant of per-VLAN STP) is the default almost everywhere today. Plain 802.1D is mostly there so you recognize the port states and timers by name. MST only earns its complexity on networks with enough VLANs that per-VLAN instances would genuinely overload the switches.