BGP based Multihoming in VPLS IETF75

1
BGP based Multi-homing in VPLSIETF-75

• Bhupesh Kothari
• Kireeti Kompella

Wim Henderickx Florin Balus
2
Statusdraft-kothari-henderickx-l2vpn-vpls-multiho
ming-01.txt

• Common BGP based multi-homing procedures for LDP
  and BGP signaled VPLS
• Draft is based on
• draft-kompella-l2vpn-vpls-multihoming-02.txt
• draft-henderickx-l2vpn-vpls-multihoming-00.txt

3
Outline

• Provisioning Model
• Multi-homing NLRI
• Designated Forwarder Election
• Multi-AS VPLS
• MAC Flush

4
Background BGP VPLS

• PW infrastructure is created based on VE-IDs (RFC
  4761)

VE-ID2
BGP VPLS
PE2
A
C
VE-ID1
PE1
PE3
B
VE-ID3

• Single VE-ID on PE1 is sufficient for customer
  sites A and B
• VE-ID is assigned per VPLS instance irrespective
  of ACs (no customer site attached to PE3)

5
Background LDP VPLS

• PW infrastructure is created based on procedures
  defined in RFC 4762

LDP VPLS
PE2
A
C
PE1
PE3
B

• PWs can be created either using FEC-128 or
  FEC-129
• Procedures defined in this draft for LDP signaled
  VPLS do not depend on use of BGP Auto-discovery

6
Provisioning Model

• Multi-homing Identifier (MH-ID) Same MH-ID must
  be configured on PEs connected to the same
  customer site

MH-ID10
VPLS
PE3
B
PE1
Customer site A dual-homed to PE devices for
redundancy
PE4
C
A
PE2
MH-ID10
7
Multi-homing NLRI

• Use of NLRI defined in RFC 4761
• VE-ID field in the NLRI is set to MH-ID
• LB, OFF and LR are set to zero

MH-ID10
VPLS
PE3
B
PE1
PE4
C
A
PE2
MH-ID10
MH-ID is not used to create PWs
8
Designated Forwarder Election
Based on DF tie-breaking rules, PE1 is DF
PE1 MH-ID10, PREF200
VPLS
PE1
PE3
B
Based on DF election, PE3 can chose to not send
traffic to PE2 (depends on whether there are
additional sites on PE2 or not)
A
PE2
AC is put in non-forwarding state
PE2 MH-ID10, PREF100
PE2 loses DF election elected as non-DF
9
Multi-AS Procedures

• For the tie-breaker, the loopback of the PE
  originating the MH NLRI is needed
• This may be overwritten in a multi-AS environment
  (such as in option B)
• To preserve this, we use the Route Origin
  Extended Community

10
Multi-AS Procedures

• Local Pref is not passed across ASes
• Again, this is needed for tie-breaking
• To solve this, the VPLS Preference field in the
  L2 info community is used
• Compatibility between LP and VPLS Pref is
  described in the document

11
Multi-AS Procedures

• For both BGP and LDP signaling, there are three
  models for inter-AS operation
• option A direct, back-to-back connections
  between ASes, with each AS treating the other as
  a CE device
• option B ASBRs mediating the VPLS
• option C multi-hop eBGP connections between
  RRs in each AS
• This is completely analogous to IP VPNs

12
MAC Flush

• If the designated forwarder (PE1) for a
  multi-homed CE goes down, remote PEs (PE3) should
  flush the MACs they learned from PE1
• Ideally, this is done implicitly
• However, if PE3 is not upgraded, it may not
  flush thus, in some cases, PE1 may have to send
  an explicit flush to the remote PEs

13
Next Steps

• Propose to adopt this as WG document?

14

• Questions/Comments?