PBB-EVPN ISID-based C-MAC-Flush

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶

This Internet-Draft will expire on 25 April 2024.¶

Copyright Notice

Copyright (c) 2023 IETF Trust and the persons identified as the document authors. All rights reserved.¶

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶

Table of Contents

1. Introduction

[RFC7623] defines how Provider Backbone Bridging (PBB) can be combined with Ethernet Virtual Private Networks (EVPN) to deploy ELAN services in very large MPLS networks. [RFC7623] also describes how Single-Active Multi-homing and per-I-SID Load-Balancing can be provided to access devices and aggregation networks. When Access Ethernet/MPLS Networks exists, [I-D.ietf-bess-evpn-virtual-eth-segment] describes how virtual Ethernet Segments (ES) can be associated to a group of Ethernet Virtual Circuits (EVCs) or even Pseudowires (PWs). In order to speed up the network convergence in case of failures on Single-Active Multi-Homed Ethernet Segments, [RFC7623] defines a Customer MAC flush mechanism that works for different Ethernet Segment B-MAC address allocation models.¶

In some cases, the administrative entities that manage the access devices or aggregation networks do not demand Multi-Homing Ethernet Segments (ES) from the PBB-EVPN provider, but simply multiple single-homed ES. Single-homed ES use null ESIs, whereas multi-homed ES use non-null ESIs. If case of using single-homed ES, the PBB-EVPN network is no longer aware of the redundancy offered by the access administrative entity. Figure 1 shows an example where the PBB-EVPN network provides four different Attachment Circuits for I-SID1, with those Attachment Circuits not being part of any Ethernet Segment or virtual Ethernet Segment (therefore they are referred to as null virtual Ethernet Segment).¶

     <----G.8032--><--PBB-EVPN Network---><----MPLS---->
          Access          MPLS                Access
           Ring                               Network
     I-SID1    ESI +------+         +------+
     +----+    null| PE1  |---------| PE3  |
     |CE1 |--------|B-MAC1|         |B-MAC3| ESI null
     +----+  active|      |         |      |----+ PW
       |           +------+         +------+     \active
       |             |                 |          \  +----+
       |             |                 |           ==|CE3 |I-SID1
       |             |                 |          /  +----+
       |stb    ESI +------+         +------+     / PW
     +----+    null| PE2  |         | PE4  |----+ standby
     |CE2 |--------|B-MAC2|         |B-MAC4| ESI null
     +----+  active|      |---------|      |
     I-SID1        +------+         +------+

In the example in Figure 1, CE1, CE2 and CE3 are attached to the same service, identified by I-SID1, in the PBB-EVPN PEs. CE1 and CE2 are connected to the PEs via G.8032 Ethernet Ring Protection Switching, and their Attachment Circuits to PE1 and PE2 are represented by a port and VLAN identifier. CE3 is dual-homed to PE3 and PE4 through an active-standby PW, and its Attachment Circuit to the PEs is represented by a PW. Each of the four PEs uses a dedicated Backbone MAC address as source MAC address (B-MAC1, B-MAC2, B-MAC3 and B-MAC4, respectively) when encapsulating customer frames in PBB packets and forwarding those PBB packets to the remote PEs as per [RFC7623]. There are no multi-homed Ethernet Segments defined in the PBB-EVPN network of the example, that is why the four Attachment Circuits in Figure 1 show the text "ESI null", which means the Ethernet Segment Identifier on those Attachment Circuits is zero. Since there are no multi-homed ES defined, the PEs keep their Attachment Circuits active as long as the physical connectivity is established and the CEs are responsible for managing the redundancy, avoiding loops and providing per-I-SID load balancing to the PBB-EVPN network. Examples of CEs managing their own redundancy are described in [G.8032], or [RFC4762] for active/standby Pseudowires.¶

For instance, in normal conditions, CE2 will block its link to CE1 and CE3 will block its forwarding path to PE4. In this situation, a failure in one of the redundant Attachment Circuits will trigger the CEs to start using their redundant paths, however those failures will not trigger any Customer MAC flush procedures in the PEs that implement [RFC7623], since the PEs are not using the PBB-EVPN multi-homing procedures. For example:¶

• if the active PW from CE3 (to PE3) fails and the failure is due to the entire PE3 node failing, then the procedures in [RFC7623] guarantee that all the remote PEs flush all the Customer MACs associated with B-MAC3 (the B-MAC of PE3). In this case, CE3 detects the fault due to the PW going operationally down.¶

• however, if the active PW from CE3 (to PE3) fails (but PE3 is still operationally up), following the procedures in [RFC7623], neither PE3 nor PE4 issue a Customer MAC flush message and therefore the remote PEs will continue pointing at PE3's Backbone MAC to reach CE3's Customer MACs, until the Customer MACs age out in the I-SID1 forwarding tables. In this case, PE3 may use any of the existing PW notifications so that CE3 switches the active PW to PE4.¶

• the same issue is exposed when the active PW from CE3 switches over from PE3 to PE4 due to a manual operation on CE3; that is, neither PE3 nor PE4 trigger any Customer MAC flush notification and the remote PEs continue sending the traffic to PE3 until the Customer MACs age out.¶

[RFC7623] provides a Customer MAC flush solution based on a shared Backbone MAC update along with the MAC Mobility extended community where the sequence number is incremented. However, the procedure is only used along with multi-homed Ethernet Segments. Even if that procedure could be used for null Ethernet Segments, as in the example of Figure 1, the [RFC7623] Customer MAC flush procedure would result in unnecessary flushing of unaffected I-SIDs on the remote PEs, and subsequent flooding of unknown unicast traffic in the network. For instance, in case CE3 switches its active PW over to PE4, a potential solution reusing the existing C-MAC Flush notifications in [RFC7623] could be for PE3 to issue an update for the MAC/IP Advertisement route of B-MAC3 with a higher sequence number. However, this update would have caused unnecessary Customer MAC flushing for all the I-SIDs attached to PE3 (supposing multiple I-SIDs in PE3), and not only I-SID1.¶

This document describes an extension of the [RFC7623] Customer MAC flush procedures, so that in the above failure example, PE3 can trigger a Customer MAC flush notification that makes PE1, PE2 and PE4 flush all the Customer MACs associated to PE3's B-MAC3 and (only) I-SID1. In order to do so, this specification introduces the encoding of the I-SID in the MAC/IP Advertisement routes advertised for the B-MACs. The Customer MAC flush procedure explained in this document is referred to as "PBB-EVPN I-SID-based C-MAC-flush" and can be used in PBB-EVPN networks with null or non-null (virtual) Ethernet Segments.¶

This specification assumes that the reader is familiar with the procedures in [RFC7623].¶

2. Solution requirements

The following requirements are followed by the C-MAC-flush solution described in this document:¶

1. The solution MUST prevent black-hole scenarios in case of failures on null ES ACs (Attachment Circuits not associated to ES, that is, their corresponding ESI is zero) when the access device/network is responsible for the redundancy.¶

2. This extension described in this document MUST work with Single-Active non-null ES and virtual ES, irrespective of the PE B-MAC address assignment (dedicated per-ES B-MAC or shared B-MAC, as in [RFC7623]).¶

3. In case of failure on the egress PE, the solution MUST provide a C-MAC-flush notification at B-MAC and I-SID granularity level.¶

4. The solution MUST provide a reliable C-MAC-flush notification in PBB-EVPN networks that use Route-Reflectors (RRs). MAC flushing needs to be provided to all affected I-SIDs in spite of the BGP flush notification messages being aggregated at the RR.¶

5. The solution MUST coexist in [RFC7623] networks where there are PEs that do not support this specification.¶

6. The solution SHOULD be enabled/disabled by an administrative option on a per-PE and per-I-SID basis (as opposed to be always enabled for all the I-SIDs).¶

3. EVPN BGP Encoding for ISID-based C-MAC-flush

The solution does not use any new BGP attributes but reuses the MAC Mobility extended community as an indication of C-MAC-flush (as in [RFC7623]) and encodes the I-SID in the Ethernet Tag field of the EVPN MAC/IP advertisement route. As a reference, Figure 2 shows the MAC Mobility extended community and the EVPN MAC/IP advertisement route that are used specified in [RFC7432] and used in this document as a C-MAC-flush notification message.¶

0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Type=0x06     | Sub-Type=0x03 |   Flags       |   Reserved=0  |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                       Sequence Number                         |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

            +---------------------------------------+
            |  Route Distinguisher                  |
            +---------------------------------------+
            |  ESI = 0                              |
            +---------------------------------------+
            |  Ethernet Tag ID = I-SID              |
            +---------------------------------------+
            |  MAC Address Length = 48              |
            +---------------------------------------+
            |  B-MAC Address                        |
            +---------------------------------------+
            |  IP Address Length = 0                |
            +---------------------------------------+
            |  MPLS Label1                          |
            +---------------------------------------+

Where:¶

• The route's route distinguisher and route targets are the ones corresponding to its EVI. Alternatively to the EVI's RT, the route MAY be tagged with an RT auto-derived from the Ethernet Tag (I-SID) instead. [RFC7623] describes how the EVPN MAC/IP Advertisement routes can be advertised along with the EVI RT or an RT that is derived from the I-SID.¶

• The Ethernet Tag encodes the I-SID for which the PE that receives the route must flush the C-MACs upon reception of the route.¶

• The MAC address field encodes the B-MAC Address for which the PE that receives the route must flush the C-MACs upon reception of the route.¶

• The MAC Mobility extended community is used as in [RFC7623], where an increment in the sequence number between two updates for the same B-MAC/I-SID will be interpreted as a C-MAC-flush notification for the corresponding B-MAC and I-SID.¶

All the other fields are set and used as defined in [RFC7623]. This document will refer to this route as the B-MAC/I-SID route, as opposed to the EVPN MAC/IP Advertisement route used in [RFC7623] that contains a specific B-MAC, and the Ethernet Tag ID set to zero. This document uses the term B-MAC/0 route to represent a B-MAC route advertised with Ethernet Tag ID = 0.¶

Note that this B-MAC/I-SID route will be accepted and reflected by any [RFC7432] RR, since no new attributes or values are used. A PE receiving the route will process the received B-MAC/I-SID update only in case of supporting the procedures described in this document.¶

4. Solution description

Figure 1 will be used in the description of the solution. CE1, CE2 and CE3 are connected to ACs associated to I-SID1, where no (Multi-Homed) Ethernet Segments have been enabled, and the ACs and PWs are in active or standby state as per Figure 1.¶

Enabling or disabling I-SID-based C-MAC-flush SHOULD be an administrative choice on the system that MAY be configured per I-SID (I-Component, Service Instance Component), as opposed to being configured for all I-SIDs. When enabled on a PE:¶

1. The PE will be able to generate B-MAC/I-SID routes as C-MAC-Flush notifications for the remote PEs.¶

2. The PE will be able to process B-MAC/I-SID routes received from remote PEs.¶

The PE MUST follow the [RFC7623] procedures for C-MAC-flush. This specification brings some additional procedures when I-SID-based C-MAC-flush is enabled.¶

This C-MAC-flush specification is described in three sets of procedures:¶

• I-SID-based C-MAC-flush activation¶

• C-MAC-flush notification generation upon AC failures¶

• C-MAC-flush process upon receiving a C-MAC-flush notification¶

5. Conclusions

The I-SID-based C-MAC-flush solution described in this document has the following benefits:¶

1. The solution solves black-hole scenarios in case of failures on null ES ACs, since the C-MAC-flush procedures are independent of the Ethernet Segment definition.¶

2. This extension can also be used with Single-Active non-null ES and virtual ES, irrespective of the PE B-MAC address assignment (dedicated per-ES B-MAC or shared B-MAC).¶

3. It provides a C-MAC-flush notification at B-MAC and I-SID granularity level, therefore flushing a minimum number of C-MACs and reducing the amount of unknown unicast flooding in the network.¶

4. It provides a reliable C-MAC-flush notification in PBB-EVPN networks that use RRs. RRs will propagate the C-MAC-flush notifications for all the affected I-SIDs and irrespective of the order in which the notifications make it to the RR.¶

5. The solution can coexist in a network with systems supporting or not supporting this specification. Non-supporting systems ignore the B-MAC/I-SID routes, however they still follow the C-MAC-flush procedures in [RFC7623].¶

6. Security Considerations

Security considerations described in [RFC7623] apply to this document.¶

In addition, this document suggests additional procedures, that can be activated on a per I-SID basis, and generate additional EVPN MAC/IP Advertisement routes in the network. The format of these additional EVPN MAC/IP Advertisement routes is backwards compatible with [RFC7623] procedures and should not create any issues on receiving PEs not following this specification, however, the additional routes may consume extra memory and processing resources on the receiving PEs. Because of that, it is RECOMMENDED to activate this feature only when necessary (when multi-homed networks or devices are attached to the PBB-EVPN PEs), and not by default in any PBB-EVPN PE.¶

7. IANA Considerations

This document requests no actions from IANA.¶

8. Acknowledgments

The authors want to thank Vinod Prabhu, Sriram Venkateswaran, Laxmi Padakanti, Ranganathan Boovaraghavan for their review and contributions.¶

9. Contributors

Authors' Addresses