rfc-7623

Provider Backbone Bridging Combined with Ethernet VPN (PBB-EVPN)

A. Sajassi (Editor), S. Salam, N. Bitar, A. Isaac, W. Henderickx
date2015-09 streamIETF wgnon working group statusPROPOSED STANDARD pages23 canonicalhttps://www.rfc-editor.org/rfc/rfc7623 doi10.17487/RFC7623
This document discusses how Ethernet Provider Backbone Bridging (PBB) can be combined with Ethernet VPN (EVPN) in order to reduce the number of BGP MAC Advertisement routes by aggregating Customer/Client MAC (C-MAC) addresses via Provider Backbone MAC (B-MAC) address, provide client MAC address mobility using C-MAC aggregation, confine the scope of C-MAC learning to only active flows, offer per-site policies, and avoid C-MAC address flushing on topology changes. The combined solution is referred to as PBB-EVPN.

Extracted elements (26)

design-rationale §3.3

C-MAC learning is confined to PE nodes actively forwarding traffic to/from those addresses via data-plane learning, avoiding the waste of RIB resources on all PE nodes in an EVI that would occur with BGP-distributed C-MAC addresses.

mpls, routing

design-rationale §3.2

C-MAC mobility is supported without re-advertising B-MAC routes because C-MAC-to-B-MAC bindings are maintained in the data plane. B-MAC addresses are fixed to PE nodes/segments, so VM mobility does not trigger BGP updates.

mpls, routing

design-rationale §6.2.1.1

For All-Active multihomed segments, all PE nodes in a redundancy group MUST share the same B-MAC address per site; otherwise remote PEs oscillate between B-MAC addresses as load balancing on the CE selects different uplinks.

mpls, routing

design-rationale §9.1

In PBB-EVPN, the number of BGP MAC Advertisement routes is proportional to the number of Ethernet segments rather than the number of hosts, potentially reducing route counts by multiple orders of magnitude compared to plain EVPN.

mpls, routing

design-rationale §9.6

No C-MAC address flushing is required for All-Active multihoming topology changes because remote PEs learn C-MAC-to-B-MAC bindings in the data plane. A link failure causes only a B-MAC route withdrawal, updating the next-hop adjacency without invalidating C-MAC entries.

mpls, routing

design-rationale §3.1

PBB is combined with EVPN to reduce BGP MAC Advertisement routes by aggregating C-MAC addresses behind B-MAC addresses, enabling MAC summarization for data center interconnect scenarios where C-MAC counts can reach millions.

mpls, routing

design-rationale §6.2.1.3

Split-horizon filtering in PBB-EVPN uses the B-MAC source address rather than the ES label used in RFC 7432, because B-MAC uniquely identifies the originating site and this function is already defined by the IEEE PBB I-Component specification.

mpls, routing

design-rationale §6.2.1.2

The B-MAC address for All-Active sites can be automatically derived by flipping the 'Locally Administered' bit of the CE's LACP System Identifier MAC address, guaranteeing network-wide uniqueness and consistent assignment across all PE nodes connected to the same CE.

mpls, routing

design-rationale §5.3

Using BGP-based Inclusive Multicast Ethernet Tag routes for per-I-SID multicast pruning is advantageous over IEEE MMRP because it scales to a large number of PEs and works with any LSP type, whereas MMRP operates only over P2P pseudowires.

mpls, routing, multicast

interoperability-note §8

PBB-EVPN interworks seamlessly with IEEE 802.1aq/802.1Qbp access networks by extending the PBB data-plane encapsulation over the MPLS backbone. B-MAC addresses of 802.1aq switches are advertised using the standard EVPN MAC/IP route advertisement defined in RFC 7432.

mpls, routing

normative-requirement §8.1 MUST

B-MAC addresses must be globally unique across all interconnected networks including PBB-EVPN and IEEE 802.1aq/802.1Qbp networks. Uniqueness for All-Active segments is achieved via operator-assigned OUI-based MAC pools.

mpls, routing

normative-requirement §5.2 MUST NOT

In the MAC/IP Advertisement route, the Ethernet Segment Identifier field must be set to either 0 (for single-homed segments or multihomed segments with per-I-SID load-balancing) or MAX-ESI (for multihomed segments with per-flow load-balancing). All other ESI values are not permitted.

mpls, routing

normative-requirement §7 MAY

PE nodes MAY implement an ARP/ND-proxy function to minimize broadcast ARP/ND traffic over the MPLS network. The PE snoops ARP/ND messages to build an IP/MAC cache and responds to ARP requests for remote hosts, flooding only on cache miss.

mpls, routing

normative-requirement §5.3 SHOULD

The Route Target for the Inclusive Multicast Ethernet Tag route SHOULD be set to the value corresponding to the EVI (analogous to B-VID), and MAY be auto-derived from the Ethernet Tag (I-SID) per the overlay procedure.

mpls, routing

normative-requirement §6.2.1.2 MUST

Whenever a B-MAC address is provisioned (manually or automatically), the PE MUST transmit a MAC Advertisement route for that B-MAC address with a downstream-assigned MPLS label that uniquely identifies the address on the advertising PE.

mpls, routing

protocol-element §6.2.2.3

C-MAC flushing on link/port failure is triggered by re-advertising the B-MAC route with the MAC Mobility extended community (Counter incremented) for the shared B-MAC model, or by withdrawing the B-MAC route for the dedicated B-MAC model.

mpls, routing

protocol-element §6.2.2.1

For per-I-SID load-balancing (Single-Active mode), each PE in a redundancy group is assigned a unique B-MAC address. Implementations MAY use either a shared B-MAC for all Single-Active segments or a dedicated B-MAC per segment; both models are interoperable.

mpls, routing

protocol-element §6.4.1

For unicast forwarding, the PE PBB-encapsulates the frame using the site's B-MAC as source address and the C-MAC-to-B-MAC binding (learned via reverse data-plane learning) as the B-MAC destination. The frame is then encapsulated with an LSP tunnel label and an EVPN label associated with the B-MAC destination.

mpls, routing

protocol-element §6.4.2

Multi-destination frames are PBB-encapsulated with the site B-MAC as source and a multicast B-MAC destination selected per I-SID as defined in IEEE 802.1Q PBB. The MPLS forwarder then uses either ingress replication over MP2P/P2P LSPs or P2MP/MP2MP tunnel LSPs.

mpls, routing, multicast

protocol-element §5.5

The ESI Label Extended Community from RFC 7432 is not used in PBB-EVPN because split-horizon filtering is performed natively using the B-MAC source address, eliminating the need for the ES label mechanism.

mpls, routing

protocol-element §5.1

The Ethernet Auto-Discovery route from RFC 7432 is not used in PBB-EVPN. PBB encapsulation provides the required indirection for C-MAC addresses via B-MAC, and ESI values 0 and MAX-ESI in the MAC Advertisement route signal that no A-D route is needed for resolution.

mpls, routing

protocol-element §5.3

The Inclusive Multicast Ethernet Tag route is used for per-I-SID multicast pruning and auto-discovery of PEs participating in a given I-SID. The Ethernet Tag field is set with the I-SID value; this BGP-based pruning supersedes MMRP and works with any LSP type (P2P, MP2P, P2MP, MP2MP).

mpls, routing, multicast

protocol-element §5.2

The MAC/IP Advertisement route carries B-MAC addresses of PE nodes rather than C-MAC addresses of end-stations. The MAC address field contains the B-MAC, the Ethernet Tag is set to 0, and the route is tagged with the RT of the associated EVI (analogous to B-VID).

mpls, routing

protocol-element §5.7

The MAC Mobility Extended Community is used with B-MAC Advertisement routes. When attached to a B-MAC route, it signals that all C-MAC addresses associated with that B-MAC across all corresponding I-SIDs must be flushed on receiving PEs.

mpls, routing

security-consideration §10

All security considerations from RFC 7432 apply directly to PBB-EVPN because PBB-EVPN reuses the RFC 7432 control plane. No new security considerations are introduced: B-MAC addresses follow RFC 7432 advertisement and processing, and C-MAC addresses follow RFC 4761 data-plane learning.

security, mpls, routing

wire-format §6.5

PBB frames are transported over MPLS with the stack: IP/MPLS Header | PBB Header | Ethernet Header | Ethernet Payload | Ethernet FCS. This is identical to classical Ethernet over MPLS except for the additional PBB header inserted between the MPLS and Ethernet layers.

mpls, routing