40 likes | 162 Views
This document discusses the challenges of scaling pseudonodes during Designated Router (DR) elections in a TRILL environment, especially when dealing with different VLANs. It highlights the possibility of cloud partitioning and the need for multiple DRBs when VLAN configurations vary. Effective communication between routers despite shared VLANs is critical, and using a pseudonode per DRB/port mitigates potential issues. The document also addresses IS-IS Hello packets and the implications of nontransitive connectivity errors, emphasizing the importance of network design to ensure reliability.
E N D
Radia Perlman Radia.Perlman@sun.com TRILL issue: Scaling of pseudonodes
DR election per VLAN • Unfortunately, because of possible bridge configuration, it is possible to have a layer 2 cloud partitioned with respect to VLAN A, but not with respect to VLAN B. So you'd need two DRBs for VLAN A, and one for VLAN B • To really know if you can talk to R1 on a cloud with VLAN A, you need to tag a Hello with that
But you don't want to have 4000 pseudonodes • So...it works to have one pseudonode for each DRB/port • If R1 is DRB for VLANs {A, B, D, H}, then create a single pseudonode • In IS-IS Hellos, same name regardless of VLAN • So...possible for R1 and R2 not to be able to talk even though they're on the same VLAN
If R2 can't talk to R3 • Link state database says next hop is pseudonode, and next hop is R3 • But R2 can't talk to R3 • Already in IS-IS: in that case, send to DR • This was put in because of bizarre error cases of nontransitive connectivity • Which bridges give us the opportunity to cause