Multiple Spanning Tree (MST): regions, instances and interoperating with PVST+

Why MST exists

PVST+ and Rapid PVST+ run one tree per VLAN. That is fine with twenty VLANs, but a data centre or campus with 500 VLANs would need 500 trees, each sending BPDUs every two seconds on every trunk. Most networks only have two or three different paths anyway. Multiple Spanning Tree (MST, IEEE 802.1s, now part of 802.1Q) lets you build just as many trees as you have useful paths and map VLANs to them. It uses rapid convergence like RSTP and is an open standard, so it suits mixed-vendor networks.

Regions and instances

An MST region is a group of switches with the same region name, revision number and VLAN-to-instance mapping. The switches compare a digest of the mapping in their BPDUs. If even one VLAN is mapped differently, they treat each other as separate regions, which usually gives a much less efficient tree.

• Instance 0 is the IST (internal spanning tree). It always exists, carries every VLAN not mapped elsewhere, and is the only instance that sends BPDUs outside the region.
• Instances 1 and up are MSTIs. Each has its own root and port roles inside the region but sends its information inside the instance 0 BPDU, so there is still only one BPDU per link.
• The CIST (common and internal spanning tree) joins the IST of every region and any non-MST switches into one tree across the whole network.

Configuring MST

spanning-tree mode mst
spanning-tree mst configuration
 name CAMPUS
 revision 3
 instance 1 vlan 10-19
 instance 2 vlan 20-29
 exit
spanning-tree mst 0-1 priority 24576
spanning-tree mst 2 priority 28672
show spanning-tree mst configuration
show spanning-tree mst 1

On the second distribution switch reverse the priorities for instances 1 and 2. Edit the configuration on every switch in exactly the same way; VTP version 3 can distribute it for you. Changing the mapping later changes the digest and can briefly split the region, so plan it in a maintenance window. A good habit is to map VLANs into instances in advance, including VLANs not yet created.

MST and PVST+ together

During a migration, or where a site meets another team's switches, an MST region may border switches running PVST+ or Rapid PVST+. The region sends and receives on its boundary ports using the IST, while the PVST+ side sends one BPDU per VLAN. Cisco switches handle this with PVST simulation: the MST boundary port copies its IST information into PVST+ BPDUs for every VLAN and checks that what comes back is consistent.

The rule that keeps this working is simple: put the root bridge for all VLANs inside the MST region, by making a region switch the root of instance 0 with a low priority. If a PVST+ switch outside becomes root for some VLANs and not others, the boundary port receives inconsistent information, and the switch blocks it as "PVST simulation inconsistent". Check for this in the logs and in show spanning-tree inconsistentports.

Real-world example: merging two networks

A housing association merged with a smaller one and connected the two head offices with a fibre link. One network ran Rapid PVST+ on Cisco switches; the other ran MST on another vendor's switches. Users in the second office lost connectivity on some VLANs after the link went live. The logs on the Cisco boundary switch showed PVST simulation inconsistencies: the Cisco core was root for VLAN 10, but an MST switch was root for everything else. The team lowered the instance 0 priority in the MST region and later moved the Cisco switches into the same MST region with a shared mapping, leaving one consistent tree.