Multicast in VPLS
draft-raggarwa-l2vpn-vpls-mcast-00.txt
Rahul Aggarwal (Juniper)
Yuji Kamite (NTT)
Luyuan Fang (AT&T)
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
1
Co-Authors
 Yakov Rekhter (Juniper)
 Chaitanya Kodeboniya (Juniper)
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
2
Agenda
 Update since last IETF
 Limitations of existing VPLS proposals for Multicast
 VPLS Multicast Architecture – where do the various drafts
fit
 Avoiding PIM snooping on the PWs
 Use of P-Multicast Trees
 Conclusion
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
3
Update Since Last IETF
 Part of the work was presented in this WG at the
last IETF when it was part of draft-raggarwa-l3vpnmvpn-vpls-mcast-01.txt
 The VPLS mechanisms are now in a separate new
draft as per WG feedback
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
4
Current VPLS proposals:
 “Virtual Private LAN Service” (draft-ietf-l2vpn-vplsbgp)
 “Virtual Private LAN Services over MPLS” (draftietf-l2vpn-vpls-ldp)
 Limitations of these proposals for VPLS
multicast…
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
5
Limitations of current VPLS proposals
for VPLS Multicast
 Do not allow the use of P-Multicast Trees for
VPLS multicast data traffic
•Desirable for optimizing bandwidth efficiency
 PEs with VPLS sites that do not have receivers
in a given multicast customer (S, G) receive
traffic for that multicast stream
 Focus on optimizing state and not bandwidth.
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
6
Design Objectives for multicast
support in VPLS service (not a
complete list)
Optimize Bandwidth:
 A given customer (multicast)
packet should traverse a given
service provider link at most
once
 Deliver customer multicast
traffic to only PEs that have
(customer) receivers for that
traffic
 Deliver customer multicast
traffic along the “optimal” paths
within the service provider (from
the ingress PE to the egress PEs)
Optimize State:
• The amount of state within the
service provider network required
to support Multicast in VPLS
service should be no greater than
what is required to support
unicast in VPLS service
• The overhead of maintaining the
state to support Multicast in VPLS
service should be no greater than
what is required to support
unicast in VPLS service
Optimizing Bandwidth and State are conflicting goals
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
7
VPLS Multicast Architecture
Control Plane
 VPLS Auto-Discovery
• Use existing VPLS auto-discovery mechanisms
 Allow elimination of flooding
• PE-CE snooping – draft-serbest-l2vpn-vpls-mcast
• In VPLS a PE does not maintain a layer 2 adjacency with a CE
• PE-PE reliable exchange of multicast control messages
• Allow avoiding PIM-IGMP snooping overhead on PWs
• PIM with reliability extensions or
• BGP
• Draft-raggarwa-l2vpn-vpls-mcast-
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
8
VPLS Multicast Architecture
Data Plane
Tree = PIM, RSVP-TE P2MP LSPs, Receiver Initiated P2MP LSPs
Separate Tree
per-set-of VPLSs
“Inclusive Mapping”
Separate Tree
per-set-of C-(S, G)s
“Selective Mapping”
Ingress
Replication
State =
Unicast VPLS
Separate Tree for
Every C-(S, G)
Aggregate State
Unbounded
State
Increasing P-router state and
Bandwidth efficiency
Decreasing P-router state and
Bandwidth efficiency
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
9
PIM snooping – implications on the
state maintenance on PE routers
 PE router has to maintain (S, G) state at least for all the (S,G) received
from all the local CEs
• E.g., assume PE with 1,000 CEs/sites, each VPLS site has at any
given point in time on average receivers for 3 groups, PE has to
maintain at least 3,000 (S,G) entries
 PE router maintains (S, G) state by processing PIM Join messages
received from (a) all sites of VPLSs connected to that PE, AND (b) all the
remote PEs that have members of these VPLSs
• E.g., assume PE router with 1,000 CEs/sites, each VPLS site has at
any given point in time on average receivers for 3 group, each group
is present on average in 10 sites, PE router has to process ~300
PIM Join per second, and ~900 (S, G) entries per second in a steady
state
• due to periodic PIM Join and PIM Join suppression
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
10
Avoiding PIM Snooping on PWs
Reliable Exchange of Multicast Control
Messages between PEs
VPLS A
Site 4
CE-A4
PE 4
VPLS A
Site 1
PIM Join
C-S, C-G –
Snooped at PE1
CE -A1
VSI-A
VRF-B
PE 1
CE-B1
BGP MVPN Routing Information Update:
<RD, C-S, C-G, Originator PE – PE1
Upstream PE – PE2, RT>
C-S -> C-G
VSI-A
RR
CE-A3
VSI-B
VPLS A
Site 3
PE 3
CE-B2
PEs have I-BGP Peering Only
With the RR
Copyright © 2004 Juniper Networks, Inc.
VPLS
B
Site 3
CE-B3
VSI-A
VSI-B
VPLS
B
Site 1
PE 2
VSI-A
CE-A2
VPLS A
Site 2
VPLS
B
Site 2
Proprietary and Confidential
www.juniper.net
11
VPLS Multicast
Data Plane Tunnels
 SP Multicast Trees
• Draft-raggarwa-l2vpn-vpls-mcast
• Aggregate Trees (Inclusive Mapping)
• Aggregate Data Trees (Selective Mapping)
• Use PE-CE snooping and PE-PE Reliable multicast message
exchange
 Ingress Replication
• Existing VPLS proposals AND
• PE-CE snooping and PE-PE Reliable multicast message
exchange
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
12
Aggregate P-Multicast Trees
 Allow one P-multicast Tree to be shared across multiple
VPLSs
 Can be setup using PIM or P2MP MPLS TE
• No architectural limitation on the P-multicast tree
technology
 Requires a MPLS label to demultiplex a particular VPN
• ‘Upstream’ label allocation by the root of the tree
• Egress PEs maintain a separate label space for each Pmulticast tree root
 State grows less than linearly with number of VPLSs
• Some efficiency of multicast routing may be sacrificed
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
13
VPLS Multicast Data Plane
P-Multicast Trees (Inclusive
Mapping)
VPLS A
Site 4
CE-A4
PE 4
VPLS A
Site 1
CE -A1
PE 2
VSI-A
VPLS
B
Site 3
CE-B3
VSI-A
VSI-A
VSI-B
PE 1
VSI-B
VPLS
B
Site 1
CE-B1
RR
CE-A3
VSI-B
VPLS A
Site 3
PE 3
CE-B2
VPLS Membership Discovery:
<Aggregation Capability> eg PE1
Aggregate Tree – PE2 as Root
C-S -> C-G
VSI-A
CE-A2
VPLS A
Site 2
VPLS
B
Site 2
BGP Signaled VPLS – Tree Binding:
<RD, Root PE, Upstream Label, RT, Tree Identifier>
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
14
Conclusion
 Comments ?
Copyright © 2004 Juniper Networks, Inc.
Proprietary and Confidential
www.juniper.net
15