PIRC Architecture July 2007 Plenary Session San Francisco

PIRC Architecture July 2007 Plenary Session San Francisco Mike Takefman Cisco Systems Mike Takefman

My PIRC High Level Goals • Built on top of 802. 17 b MAC • Assumption is an L 2 switched network • PIRC has to provide loop-free dual connectivity • Multiple Load-Balancing methods • Compatible with non-PIRC stations • ideally no MAC or operational changes needed for other nodes • usable with existing standard product chips 7/16/2007 IEEE 802. 17 RPRWG Mike Takefman

Where To Put Filtering • PIRC is essentially adding new filtering rules to the system • assumption is that ingress and egress filter rules exist • Our scope requires that these filtering rules exist in the MAC level, but the actual position is implementation dependant • side effect on whether 802. 17 b learning occurs • no easy way to arbitrarily allow or not allow learning • question is one of compliance • Tributary support requires egress filtering rules in addition to ingress rules 7/16/2007 IEEE 802. 17 RPRWG Mike Takefman

Hash Filtering • Previous presentations have pointed out that certain constraints must be followed in order to insure connectivity • net result is that the basic hash function is based on outer VLAN • hence this form of hash filtering is essentially VLAN filtering and any other hashing function is by definition proprietary 7/16/2007 IEEE 802. 17 RPRWG Mike Takefman

2 Span Failure Scenarios • It has been suggested that PIRC support both 1 and 2 span failures • it must be assumed that the network is otherwise loop free • 2 cuts on a leaf ring is trivial as a loop cannot be created C D • 2 cuts on an intermediate ring is more interesting and has a few cases A 7/16/2007 IEEE 802. 17 RPRWG B Mike Takefman

2 Span Failure Scenarios • Isolated Node • no connectivity to that node • no loops created • Nothing can be done, and there are no PIRC issues C A 7/16/2007 IEEE 802. 17 RPRWG D B Mike Takefman

2 Span Failure Scenarios • Isolated Neighbour Ring • no connectivity to that sub-tree • no loop created • Nothing can be done, and there are no PIRC issues C A 7/16/2007 IEEE 802. 17 RPRWG D B Mike Takefman

2 Span Failure Scenarios • Isolated Partial Ring • full connectivity possible • no loop created • PIRC could restore connectivity • A&B operated normally • D&C both do full forwarding C A 7/16/2007 IEEE 802. 17 RPRWG D B Mike Takefman

2 Span Failure Scenarios • Split Ring • full connectivity possible BUT • loop created with current suggested PIRC filtering • PIRC could restore connectivity • A&B, C&D forward everything from neighbour ring to “broken ring” • one of A|B, C|D forward to neighbour ring • never forward back from mate • done with egress filtering if tribs are needed else ingress 7/16/2007 IEEE 802. 17 RPRWG C A D B Mike Takefman

2 Span Failure Scenarios • How to recognize split ring? Nodes on affected ring observe • mate has disappeared on one ring but not the other C A 7/16/2007 IEEE 802. 17 RPRWG D B Mike Takefman

2 Span Failure Scenarios • Does the split ring scenario scale to 3 PIRC pairs? • A&B, C&D are partially isolated from B&E • A&B is split from C&D • the problem is that the split rules are different from the partially isolated rules • implies you need to apply rules based on ring. SA -> YUCK 7/16/2007 IEEE 802. 17 RPRWG D C B E A F Mike Takefman

Conclusions • Dual Failure scenario can only be handled for limited cases • likely not worth doing 7/16/2007 IEEE 802. 17 RPRWG Mike Takefman