[c-nsp] MST configuration - VLAN "islands"

[Grzegorz Banasiak]

Hello,

Let's assume the following topology:

                          [X]
        Aggr switch A  ---------  Aggr switch B
              |                         |
              | [X]                     | [X]
              |          [X,Y]          |
       Access switch 1A  ----- Access switch 1B
              |                         |
              | [X,Y]                   | [X,Y]
              |                         |
       Access switch 1C  ----- Access switch 1D
                         [X,Y]

A set of 4 access switches is connected to a pair of aggregation switches.
Access switches carry VLANs X and Y. VLAN Y has a local "scope" - only VLAN X
needs to be passed to the aggregation layer.

I'm looking for a decent MST (Multiple Spanning Tree) configuration guidelines
for such a case assuming that a similar set of 4 access switches is replicated
many times. Each time a set is connected to the same pair of aggregation
switches, only the VLAN IDs differ - (X', Y'), (X'', Y''), (X''', Y'''), etc.

Here's my approach:
- all switches in the same MST region - simplicity and fast convergence (MST
falls back into 802.1D on a region boundary, right?),
- instance 0 is left empty (no VLANs attached), aggregation switches A and B are
primary/secondary IST root bridges,
- VLANs X, X', X'', etc. are assigned to instance 1 for which aggregation
switches A and B are primary and secondary root bridges,
- VLAN Y is assigned to instance 2 for which 2 of access switches 1A/B/C/D are
configured as primary/secondary root - this gives me the opportunity to enable
different links for X and Y,
- VLAN Y' is assigned to instance 3,
- VLAN Y'' is assigned to instance 4.
...

In this way I can have 14 sets of access switches due to the limit of 15
instances (excluding instance 0) in MST.

Questions:
- Does it make sense?
- Is it possible to reuse instances for "local scope" VLANs Y', Y'', Y'''?
- Is it really necessary to configure all switches in the MST region with
exactly the same instance to VLAN mappings even if a particular switch does not
carry some of the VLANs in the mapping? For example access switches 2A/B/C/D do
not carry VLANs X,Y from access switches 1A/B/C/D. Cisco configuration guides
say "Yes, it is neccessary." but I'm asking about real life experience here.
- Has the physical topology got any fundamental flaw?

And no, I cannot switch to RPVST+ ;)

Regards.