pvst+ mono spanning tree (MST) regions and data frames flow

[adimnewtemp]

Hi,

As far as i understand when there is a PVST+ region (which means that there are ISL links between the switches) and a MST(mono spanning tree) region, VLAN's 1 STP instance of PVSTP+ region is mapped to the single CST unique instance of MST region and vica versa.

My question concerns what happens with data frames belonging to VLANs other than vlan 1. Since the PVST+ BPDUs other than VLAN 1 are tunneled into the MST region, this means that their STP instance blocks somewhere into the PVST+ region

So in case that CST instance blocks into the MST region and the trunks into the MST region are 802.1q, what happens with data frames other than VLAN's 1, since for them there are 2 STP blocks: one inside the CST region (due to CST instance) and another one inside the PVST+ region (due to corresponding's VLAN PVST+ STP instance)? What i point to is that if there are workstations into the MST region belonging to VLANs other than 1 then intercommunication with ALL corresponding workstations of same VLAN of the PVST+ region is not possible.

Please point out any misunderstanding

Thanks in advance

[Peter Paluch]

Hi,

My favourite STP topics ;)

I am afraid, though, that you are confusing the terms and as a result, I am not entirely sure about the particular setup you are asking about. You have used the MST acronym and explained it as Mono Spanning Tree.

In short, there is no such term as Mono Spanning Tree. Very old Cisco documents have been using this term to denote a network region running legacy STP without any per-VLAN semantics. However, nowadays, the term MST is used exclusively to refer to 802.1s (now 802.1Q) Multiple Spanning Tree. My confusion is compunded by the fact that you are referring to the CST - the Common Spanning Tree - which is the current standardized name for the legacy STP region and, at the same time, for a specific type of tree in Multiple Spanning Tree, and the CST term did not exist in times when Mono Spanning Tree term was in use.

So before we can answer your question, let's clear up the terminology: Are you interested about the interoperability issues between PVST+ and legacy STP regions, or between PVST+ and Multiple Spanning Tree regions?

Best regards,
Peter

[adimnewtemp]

Hi,

Thanks your your quick reply.

What i am interesting about is  between PVST+ and legacy STP regions.

Waiting your comments

[adimnewtemp]

Hi Peter,

what i also mean by 802.1q trunks are trunks that carry all possible Vlans with the normal IEEE 802.1q tag scheme

thanks

[Peter Paluch · Accepted solution]

Hi,

Okay, interoperability between a PVST+ and a legacy STP region, got it.

Let's assume that the PVST+ region is connected to the STP region using a trunk link. On trunks, PVST+ ports send out both STP BPDUs and PVST+ BPDUs. The PVST+ BPDUs are formatted differently (different destination MAC address, SNAP encapsulation instead of LLC) and are thus not interpreted by STP switches; in fact, they are just flooded as ordinary multicasts. Some people call this "tunnelling PVST BPDUs through STP regions" but there is no tunnelling involved - these BPDUs are simply flooded just like any unknown multicasts.

STP BPDUs sent from the PVST+ port are always derived from VLAN1 PVST+ instance. Also, STP BPDUs received on a PVST+ port are processed in VLAN1 PVST+ instance. In effect, the PVST+ VLAN1 instance and the legacy STP region merge together, resulting in a single loop-free topology. In the STP region, this loop-free topology applies to all VLANs indiscriminately. In the PVST+ region, this loop-free topology applies to VLAN1 only. As a result, from the viewpoint of the PVST+ region in all other VLANs, the entire STP region reduces to a single, non-looped, shared segment. It is as if you interconnected all PVST+ ports to the STP region by a single hub. On this topology, the PVST+ instances resolve the remaining loops in VLANs 2-4094 (remember that VLAN1 is already loop-free).

Since the PVST+ BPDUs other than VLAN 1 are tunneled into the MST STP region, this means that their STP instance blocks somewhere into the PVST+ region

Yes, this is true. To VLANs other than VLAN1, the STP region is just a shared segment.

So in case that CST instance blocks into the MST region and the trunks into the MST region are 802.1q, what happens with data frames other than VLAN's 1, since for them there are 2 STP blocks: one inside the CST region (due to CST instance) and another one inside the PVST+ region (due to corresponding's VLAN PVST+ STP instance)?

No VLAN has 2 STP blocks. Recall that if there is a block in the STP region, then it applies to all VLANs. No VLAN can loop through the STP region, not even the PVST+ BPDUs. As a result, the PVST+ instances in the PVST+ region have no reason to cause another STP block in the PVST+ region because the loop has already been cut in the STP region.

Let me put it differently: Either the STP region has somewhere a total STP block, resulting into partitioning it into several STP subregions interconnected by the PVST+ region (in which case this loop resolution affects all VLANs and thus all PVST+ instances), or the STP region remains contiguous - albeit loop-free - and then the PVST+ instances treat it as a shared segment and resolve the loops in the PVST+ region. Does this make sense?

Best regards,
Peter

[adimnewtemp]

"Either the STP region has somewhere a total STP block, resulting into partitioning it into several STP subregions interconnected by the PVST+ region (in which case this loop resolution affects all VLANs and thus all PVST+ instances)"

Peter,

From my point of view if the STP region has somewhere a block, then there block all vlans data frames but does NOT the multicast frames of the BPDUs of the PVST+ region. So these multicast frames which each one corresponds to one STP instance of a VLAN according to the PVST+ meaning, are not blocked there; so their instance blocks somewhere else; so there are 2 blocks(the one is that of CST) for each vlan,each one in each region

thanks.

[adimnewtemp]

Peter,

Anyway what i understand that you mean, is that if there is a CST block into the legacy region, then nothing will pass; indepedendly of if that would be a frame with a multicast address.Please correct me if  i am wrong

Thanks in advance

[Peter Paluch · Accepted solution]

Hello,

From my point of view if the STP region has somewhere a block, then there block all vlans data frames but does NOT the multicast frames of the BPDUs of the PVST+ region.

This assumption is incorrect. If the STP region has an STP block somewhere then absolutely nothing will pass - neither unicasts nor broadcasts nor multicasts nor PVST+ BPDUs (which are, to non-Cisco switches, just ordinary multicasts). There is absolutely no reason why multicasts would pass an STP block. If that was true, STP would be useless, as multicasts could still be caught in a loop even though STP constructed a loop-free topology.

Best regards,
Peter

[adimnewtemp]

Peter,

Thanks for addressing my confusion]

Best Regards

thanasis

[Peter Paluch]

Thanasis,

It has been a pleasure. You are always welcome.

Best regards,
Peter