| Internet-Draft | BIER Prefix Redistribute | December 2021 |
| Zhang, et al. | Expires 26 June 2022 | [Page] |
This document defines a BIER proxy function to support a single BIER sub-domain over multiple underlay routing protocol regions (Autonomous Systems or IGP areas). A new BIER proxy range sub-TLV is defined to redistribute BIER BFR-id information across the routing regions.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC2119.¶
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 26 June 2022.¶
Copyright (c) 2021 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.¶
It is assumed that readers of this document are familiar with BIER architecture [RFC8279] and OSPF/ISIS/BGP signaling for BIER [RFC8401], [RFC8444], [I-D.ietf-bier-idr-extensions], [I-D.ietf-bier-ospfv3-extensions]. The following terminologies are listed for convenience. (to be added)...¶
BIER [RFC8279] is a new multicast architecture that does not need per-tree state inside a network for multicast forwarding. BIER forwarding state (which is not per-tree) is built according to a routing underlay, which is defaulted to an IGP domain though not limited to that. In this routin underlay, BIER information like BIER-prefix, BFR-id, subdomain-id, and encapsulation is propagated using IGP or BGP as specified in documents such as [RFC8401], [RFC8444], [I-D.ietf-bier-idr-extensions], [I-D.ietf-bier-ospfv3-extensions], [I-D.ietf-bier-lsr-ethernet-extensions].¶
In some deployment situations, different routing protocols may be used in differet parts of a network yet there are just small number of BFERs in each protocol domain, so a single BIER sub-domain is desired. This requires BIER information redistribution among different protocols, as described in the following two sections.¶
+----+ +----+
+----+ R1 +-------------+ R2 +-----+
| +----+ +----+ |
| OSPF / ISIS |
| domain 1 |
| |
| +----+ +----+ |
+-----+ +------------+ +-----+
+------------+ R3 +---+ +---+ R4 +------------+
| +----+ | | +----+ |
| OSPF | | OSPF |
| | | |
| domain 2 | | domain 3 |
| +----+ +----+ | | +----+ +----+ |
+--+ Rm +-----+ Rn +--+ +---+ Rx +----+ Ry +--+
+----+ +----+ +----+ +----+
Figure 1
¶
While one could treat each IGP domain in the above network as a separate BIER sub-domain, the border routers R3/R4 would need decapsulate incoming BIER header in one BIER sub-domain, forward based on flow overlay per-tree state, and re-ecapsulate with a BIER header for forwarding in the next BIER sub-domain. This not only is inefficient in forwarding, but also require per-tree state on the border routers, which is undesired.¶
A better solution is to treat the entire network of multiple IGP domains as single BIER sub-domain with a single routing underlay.¶
Figure 1 could also depict a Seamless MPLS [I-D.ietf-mpls-seamless-mpls] network, where BGP-LU [RFC8277] is used to distribute routes for edge routers (e.g. R1/R2/Rm/Rn/Rx/Ry) among the edge routers and border routers (e.g. R3/R4), but those routes are not redistributed into other IGP areas/domains. With that, internal routers in an area/domain will only have routes to local nodes, yet they still need to build BIER forwarding state for BFERs in other areas/domains.¶
For the multiple IGP domains scenario in Section 2.1, BIER information from one domain needs to be redistributed into another domain, like that BIER information is redistributed from one IGP area to another as specified in [RFC8401], [RFC8444], and [I-D.ietf-bier-ospfv3-extensions].¶
Specifically, when an ASBR redistributes BIER prefixes for BFERs from one protocol domain to another, BIER information is also redistributed except the encapsulation information (because BFRs in one domain will not directly send BIER packets to BFRs in the other domain so only the BFR-IDs of the BFERs matters). When BIER prefixes for non-BFIR/BFER (i.e. whose BFR-ID is 0) are redistributed, BIER information is not redistributed.¶
If route summarization is used, because a summarized prefix may cover many BFERs, the BFR-IDs of those covered BFERs needs to be explicitly listed in proxy range sub-TLV (see Section 4.2). In case of Seamless MPLS (Section 2.2), when a border router advertise BIER information for itself in one area/domain, it also explicitly lists the BFIRs/BFERs in other areas/domains that are reachable via itself in the proxy range sub-TLV.¶
In figure 1, R3/R4 connects two routing domains. After R3 receives BIER information for Rm/Rn from domain 2 and redistribute to domain 1, BFRs in domain 1 can build BIER forwarding state for BFERs in domain 2 through R3. Similarly, R3 receives BIER information for R1/R2 from domain 1 and redistribute into domain 2. BFRs in domain 2 can build BIER forwarding state for BFERs in domain 1 through R3.¶
For example, in this network, suppose that Rm and Rn have the prefix of 201.1.1.1/32, 201.1.1.2/32. In order to build one BIER sub-domain which includes these three IGP domains, R3 advertises the BFR-ids of Rm/Rn with associated prefixes (201.1.1.1/32, 201.1.1.2/32) into the upper domain. Similarly, R4 advertises the BFR-ids of Rx/Ry with associated prefixes (202.1.1.1/32, 202.1.1.2/32) into the upper domain too.¶
And R3/R4 advertises the prefixes of R1 and R2 (suppose that the prefixes are 200.1.1.1/32 and 200.1.1.2/32) with associated BFR-ids into IGP domain 1 and domain 2. Also, R3 advertises the prefixes learned from R4 (202.1.1.1/32, 202.1.1.2/32) with associated BFR-ids into IGP domain 1. R4 also advertises the prefixes (201.1.1.1/32, 201.1.1.2/32) with associated BFR-ids into IGP domain 2.¶
Obviously, in order to build the large BIER sub-domain, the BFR-id of edge router in each IGP domain MUST NOT overlap.¶
Consider a BIER sub-domain that spans multiple routing domains. The procedures in this section apply if a border router, which is also a BFR, redistribute routing information from one routing domain into another.¶
If a redistributed route is for a host route for a BFIR/BFER (i.e. the BFR-ID is not zero) in the same sub-domain, BIER information for the BFIR/BFERs MUST be advertised in the target routing domain as following:¶
The BIER Sub-TLV (in case of OSPF2 and OSPFv3), BIER Info Sub-TLV (in case of IS-IS) and BIER TLV (in case of BGP) are encoded as specified in [RFC8444], [I-D.ietf-bier-ospfv3-extensions], and [RFC8401], and [I-D.ietf-bier-idr-extensions]. The encapsulation sub-TLV SHOULD NOT be included because it would not be used.¶
If a redistributed route is for a host route for a transit BFR (i.e. the BFR-ID is zero), BIER information for the BFR SHOULD NOT be redistributed, because it would not be used.¶
If the redistributed route is a summary or default route, and it covers some BFIR/BFERs, a BIER Sub-TLV (for OSPFv2 and OSPFv3), or a BIER Info Sub-TLV (in case of IS-IS), or a BIER TLV (in case of BGP) MUST be used to advertise the covered BFIRs/BFERs via the BIER proxy range sub-TLV as specified in the following section.¶
The BIER Sub-TLV can include a proxy range sub-TLV, which lists BFIRs/BFERs covered by a summary/default route or reachable via a BFR.¶
0 1 2 3
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 | Length | subdomain-id | resv |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BFR-id | BFR-id range |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| ... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| BFR-id | BFR-id range |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
figure 2
¶
The BIER proxy range sub-TLV is included in the BIER Sub-TLV for an aggregated/summary route prefix or default route prefix, or in the BIER Sub-TLV for a BIER prefix (for a border router in a Seamless MPLS network). Multiple BIER proxy range sub-TLVs MAY be included in the BIER Sub-TLV.¶
The range in the BIER proxy range sub-TLV can be as granular as to advertise individual BFR-ids. Though a larger range can increase advertisement efficiency, that requires careful planning for BFR-id assignment.¶
When the proxy range sub-TLV is used, the mapping between a BIER prefix and its BFR-id is no longer conveyed in the routing underlay. As a result, the mapping must be provided by other means, e.g. in the multicast overlay.¶
If a BFR receives a BIER prefix whose BIER Sub-TLV includes a proxy range sub-TLV (i.e., the Seamless MPLS scenario), it treats as if that the originator advertised a default route with the proxy range sub-TLV. Note that this imaginary default route is only for the purpose of building BIRT/BIFT entries and not used for unicast forwarding.¶
With the BIER prefixes originated in the local routing area/domain, the BIER prefixes and summary/default routes redistributed into the local routing area/domain, and the imaginary default route mentioned above, a BFR builds BIRTs as specified in [RFC8279] with entries including host/summary/default prefixes.¶
BIFT entries are then derived from a corresponding BIRT. For a BFER covered by the proxy range sub-TLVs associated with the summary/default prefixes (whether or not the deafult prefix is the imaginary one as mentioned above), its BIFT entry is derived from the summary/default prefix entry in the BIRT. It is possible that a BFER is covered in the proxy range sub-TLV of multiple default/summary routes. In that case, ECMP MAY be used and longest match SHOULD be used. For example, if ABR1 advertises default/summary route P1 while ABR2 advertises a more specific summary route P2 and both have a proxy range sub-TLV that covers BFR-ID 100, then the BIFT entry for BFR-ID 100 should follow the P2 route in BIRT.¶
With this scheme, even though the BIER prefixes are not advertised into the IGP for an area/domain in a Seamless MPLS network and unicast traffic for those BIER prefixes are tunneled through, corresponding BIFT entries are maintained inside the area/domain for the purpose of efficient BIER forwarding. Otherwise, BIER forwarding through the area/domain would be tunneled just like unicast case.¶
IANA is requested to set up a new types of sub-TLV (TLV) registry value for BIER proxy range advertisement in OSPF, ISIS, BGP, etc.¶
Implementations must assure that malformed TLV and Sub-TLV permutations do not result in errors which cause hard protocol failures.¶
The authors would like to thank Stig Venaas for his valuable comments and suggestions.¶
This appendix is to make the function understood more easily. Except for inter-area case, the function is also suitable for inter-as case. In the same example of figure 1, in case there are 40 edge routers in domain 1, the BFR-ids of domain 1 start from 51 to 90, and the prefixes of these routers start from 201.1.1.1/32 to 201.1.1.40/32. These assigned BFR-IDs are not overlapped with the BFR-IDs in any other domain.¶
In order to build a BIER sub-domain across these areas, the two advertisement methods defined in Section 4.2 can be used:¶
Then the router in the uppler domain can build the BIER forwarding table, the nexthops of BFR-IDs in the proxy range sub-TLV are set to R3.¶
The same function is also applied to R4 when it advertises the BFR-IDs to the upper domain. This method is also applied to R3/R4 when it advertises the BFR-IDs to the lower domain. When R3 advertises the prefixes from the upper domain and domain 2 into domain 1, except the host prefix of R3 with BFR-ID set to 0, R3 may advertise only one default route (0.0.0.0/0) into domain 1 if one or more continuous BFR-id range can be attached. Suppose that the BFR-id in the upper domain starts from 1001 to 1050, the BFR-id in domain 2 starts from 201 to 250, and these ranges are not overlapped with the ranges in any other domain. Suppose that the sub-domain ID is 1, the BIER proxy range sub-TLV may be advertised like this:¶
0 1 2 3
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| TBD | 2 | 1 | 0 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 1001 | 50 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 201 | 50 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
figure 3
¶
Then the BIER overlay is built among R1, R2, Rm, Rn, Rx and Ry. R3 and R4 need not maintain the multicast overlay states. The optimized summary/ aggregated or default prefix can be generated by the operation policy which is configured by the network administrator.¶
If two or more ABRs in one domain are used to reistribute the prefix, for example in figure 4 below:¶
+-------------------------------------------------------+
| +----+ +----+ BIER |
| +-----------+ R1 +-------------+ R2 +-----+ domain 1 |
| | +----+ +----+ | |
| | OSPF/ISIS | |
| | | |
| | +----+ +----+ +----+ | |
| +--+ +----+ +------------+ +-----+ |
| +--+ R5 +----+ R3 +---+ +---+ R4 +------------+ |
| | +----+ +----+ | | +----+ | |
| | | | | |
| | OSPF domain 1 | | OSPF domain 2 | |
| | | | | |
| | +----+ +----+ | | +----+ +----+ | |
| +--+ Rm +-----+ Rn +--+ +---+ Rx +----+ Ry +--+ |
| +----+ +----+ +----+ +----+ |
+-------------------------------------------------------+
Figure 4
¶
As the ABRs, except the host prefixes of R3 and R5 advertisement, R3 and R5 can both advertise the proxy range sub-TLV with their host prefix, the routers can select one of them as the nexthop, or select both of them for ECMP.¶
If R3 and R5 are allowed to advertise the summary prefix received from the upper domain. They can advertise the same summary prefixes or the different prefixes according to the operation policy. When they advertise the same summary prefixes, the R3 and R5 can also be used for ECMP. When they advertise the different summary prefixes, the more specific prefixes are used to generate the BIER forwarding table. Whatever the same or different prefixes are advertised, the nexthop is set to R3/R5.¶
In case the range of BFR-ids in one domain is overlapped with the BFR-ids in any other domain, the proxy range sub-TLV may not be used. In the same example above, if the BFR-ids in domain 1 are 21, 31, 41, etc., the BFR-ids in domain 2 are 22, 32, 42, etc., even if the summarized prefixes are not overlapped with the prefixes in any other domain, when R3 advertises the summarized prefixes in domain 1 into the upper domain, the proxy range sub-TLV may not optimize the advertisement.¶
After the forwarding plane is built, the nexthop of the range BFR-IDs is set to the ABR router. For example, when R1 receives multicast packet, and the receivers of this flow are connected by Rm and Rx, R1 encapsulates BIER header in front of the flow packet with BFR-ids set to Rm and Rx. The routers in the upper domain forward the packet to the ABR routers: R3, R4 and R5. When R3/R4/R5 receives the packet, R3/R4/R5 needs not decapsulate and re-encapsulate the packet. R3/R4/R5 just forwards the packet according to the BIER forwarding table. Similar with the upper domain, the routers in lower domain forward the packet to the edge routers: Rm and Rx. When the packet reaches Rm and Rx, Rm and Rx remove the BIER header and forward it to receivers.¶