<?xml version="1.0" encoding="US-ASCII"?>

<?xml-model href="rfc7991bis.rnc"?>

<!-- <?xml-stylesheet type="text/xsl" href="rfc2629.xslt" ?> --> 
<!-- This third-party XSLT can be enabled for direct transformations in XML processors, including most browsers -->


<!DOCTYPE rfc [
  <!ENTITY nbsp    "&#160;">
  <!ENTITY zwsp   "&#8203;">
  <!ENTITY nbhy   "&#8209;">
  <!ENTITY wj     "&#8288;">
]>
<!-- If further character entities are required then they should be added to the DOCTYPE above.
     Use of an external entity file is not recommended. -->

<?rfc strict="yes" ?>
<!-- give errors regarding ID-nits and DTD validation -->

<?rfc compact="yes" ?>
<!-- do not start each main section on a new page -->
<?rfc subcompact="no" ?>
<!-- keep one blank line between list items -->

  <rfc category="info"
	xmlns:xi="http://www.w3.org/2001/XInclude"
	docName="draft-ietf-bess-evpn-modes-interop-03"
	consensus="true"
	submissionType="IETF"
	ipr="trust200902"
	tocInclude="true"
	tocDepth="4"
	symRefs="true"
	sortRefs="true">

<front>
	<title abbrev="EVPN Interoperability Modes">EVPN Interoperability Modes</title>
    <seriesInfo name="Internet-Draft" value="draft-ietf-bess-evpn-modes-interop-03"/>

	<author fullname="Lukas Krattiger" initials="L." surname="Krattiger">
		<organization>Cisco</organization>
		<address>
			<postal>
				<street>170 W. Tasman Drive</street>
				<street/>
				<city>San Jose</city>
				<code>95134</code>
				<region>CA</region>
				<country>USA</country>
			</postal>
			<email>lkrattig@cisco.com</email>
		</address>
	</author>

	<author fullname="Ali Sajassi" initials="A." surname="Sajassi">
		<organization>Cisco</organization>
		<address>
			<postal>
				<street>170 W. Tasman Drive</street>
				<street/>
				<city>San Jose</city>
				<code>95134</code>
				<region>CA</region>
				<country>USA</country>
			</postal>
			<email>sajassi@cisco.com</email>
		</address>
	</author>

	<author fullname="Samir Thoria" initials="S." surname="Thoria">
		<organization>Cisco</organization>
		<address>
			<postal>
				<street>170 W. Tasman Drive</street>
				<street/>
				<city>San Jose</city>
				<code>95134</code>
				<region>CA</region>
				<country>USA</country>
			</postal>
			<email>sthoria@cisco.com</email>
		</address>
	</author>

	<author fullname="Jorge Rabadan" initials="J." surname="Rabadan">
		<organization>Nokia</organization>
		<address>
			<postal>
				<street>520 Almanor Avenue</street>
				<street/>
				<city>Sunnyvale</city>
				<code>94085</code>
				<region>CA</region>
				<country>USA</country>
			</postal>
			<email>jorge.rabadan@nokia.com</email>
		</address>
	</author>

	<author fullname="John E. Drake" initials="J." surname="Drake">
		<organization>Juniper</organization>
		<address>
			<email>drake@juniper.net</email>
		</address>
	</author>

	<date year="2026" />

	<!-- Meta-data Declarations -->
   <area>Routing</area>
   <workgroup>BESS Working Group</workgroup>

	<abstract>
		<t>
			Ethernet VPN (EVPN) provides different functional modes in the area
			of Service Interface, Integrated Route and Bridge (IRB) and IRB Core
			connectivity. This document specifies how the different EVPN
			functional modes and types can interoperate with each other.
			This document does not redefine the existing functional modes but
			describes how these modes interoperate.
		</t>
    </abstract>
    
    <note title="Requirements Language">
      <t>The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
      NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED",
      "MAY", and "OPTIONAL" in this document are to be interpreted as
      described in BCP 14 <xref target="RFC2119"/> <xref target="RFC8174"/> when,
      and only when, they appear in all capitals, as shown here. 
     </t>
   </note>

</front>
  
  
<middle>
	<section title="Introduction" anchor="intro">
		<t>
			Ethernet VPN (EVPN) provides different functional modes in the area
			of Service Interface, Integrated Route and Bridge (IRB) and IRB
			connection model. It is understood that the different modes are
			defined with different use-cases in mind. Even with the specific use-
			cases and the resulting mode definition, the aim of interoperability
			is critical.                                                         
			The following EVPN modes are considered for interoperability. It is
			limited to the most pertinent interop modes as opposed to all
			permutations.
		</t>

		<ul>
			<li>For Service Interfaces: VLAN-Aware Bundle and VLAN-Based.</li>
			<li>For Integrated Routing and Bridging (IRB): Asymmetric IRB and Symmetric IRB.</li>
			<li>For IRB connectivity: interface-less and interface-ful unnumbered IRB.</li>
		</ul>
		<t>
			Future revisions of this Internet-Draft might address further
			variations of interoperability.
		</t>
	</section>
  
	<section title="Valid Combinations for Interoperability" anchor="ValCom">
      	<t>
	      	The tables below provide an overview of the valid combinations for
   			interoperability described in this Internet-Draft.
   		</t>
      	<t>
	      	For the Service Interface Types as described in <xref target="RFC7432"/> section 6
   			and <xref target="RFC8365"/> section 5.1.2. Interoperability considerations are
   			provided for the VLAN-Based Service interface (<xref target="RFC7432"/>, section
   			6.1) and the VLAN-Aware Bundle Service Interface type (<xref target="RFC7432"/>
   			section 6.3). The VLAN Bundle Service Interface (<xref target="RFC7432"/> section
   			6.2) is not considered at this time.
   		</t>

		<t>
			<figure>
				<name>Service Interface Type Interoperability</name>
				<artwork><![CDATA[
+-------------------+------------+-------------+--------------------+
|                   | VLAN-Based | VLAN Bundle | VLAN-Aware Bundle  |
+-------------------+------------+-------------+--------------------+
| VLAN-Based        |    YES     |    NO       |    YES             |
+-------------------+------------+-------------+--------------------+
| VLAN Bundle       |    NO      |    YES      |    NO              |
+-------------------+------------+-------------+--------------------+
| VLAN-Aware Bundle |    YES     |    NO       |    YES             |
+-------------------+------------+-------------+--------------------+
]]>
				</artwork>
			</figure>
</t>
   
		<t>With regards to Integrated Route and Bridge (IRB), two different modes
		are defined in <xref target="RFC9135"/>, with section 5 describing Symmetric IRB and
		section 6 Asymmetric IRB:</t>
		<t>The interoperability considerations between both IRB modes,
		Asymmetric IRB and Symmetric IRB, are represented within this
		Internet-Draft.</t>
		<t>For the IRB Core Connectivity, from all the available modes defined
		in <xref target="RFC9136"/>, considered for interoperability is the interface-less
		mode (section 4.4.1) in conjunction with only one of the interface-
		ful modes, namely interface-ful IP-VRF-to-IP-VRF with Unnumbered SBD
		IRB (section 4.4.3). The close functional approximation between the two
		interface-ful modes, considerations for interoperability between
		interface-less and interface-ful Numbered are currently not
		considered. Similarly, the interoperability between the two
		interface-ful modes is currently not being considered, given the
		close functional relation and to limit permutations. Future revisions
		of this Internet-Draft might address further variations of
		interoperability.</t>

		<t>
			<figure>
				<name>IRB Core Connectivity interoperability</name>
				<artwork><![CDATA[
+-----------------+----------------+---------------+----------------+
|                 | Interface-Less | Interface-Ful | Interface-Ful  |
|                 |                | Numbered IRB  | Unnumbered IRB |
+-----------------+----------------+---------------+----------------+
| Interface-Less  |      YES       |      NO       |      YES       |
+-----------------+----------------+---------------+----------------+
| Interface-Ful   |      NO        |      YES      |      NO        |
| Numbered IRB    |                |               |                |
+-----------------+----------------+---------------+----------------+
| Interface-Ful   |      YES       |      NO       |      YES       |
| Unnumbered IRB  |                |               |                |
+-----------------+----------------+---------------+----------------+
]]>
				</artwork>
			</figure>
		</t>
	</section>

    <section title="Service Interface Interoperability" anchor="ServiceIntf">
		<section title="VLAN-Aware Bundle and VLAN-Based" anchor="VlanBund">

      		<t>
	      		<xref target="RFC7432"/> section 6 describes three different Service Interface
   				Types. The two modes in focus for interoperability are namely the
   				VLAN-Based Service Interface as defined in <xref target="RFC7432"/> section 6.1 and
   				the VLAN-Aware Bundle Service Interface as defined in <xref target="RFC7432"/>
   				section 6.3. The VLAN Bundle Service Interface is not considered.
   			</t>
      
      		<t>
	      		The VLAN-Based Service Interface defines an EVPN instance consisting
   				of only a single broadcast domain, or Single Broadcast Domain per
   				EVI, as defined in <xref target="RFC8365"/> section 5.1.2 Option 1. In this mode,
   				individual BGP Route Distinguisher (RD) and Route Target (RT) are
   				required for each EVI. Each EVI corresponds to a single MAC-VRF
   				identified by the RT. This mode has the advantage of a BGP RT
   				constraint mechanism in order to limit the propagation and import of
   				routes to only the PE that are interested. With VLAN-Based, the MAC-
   				VRF corresponds to only a single bridge table. The VLAN-Based Service
   				Interface uses the EVPN MAC/IP Advertisement route (<xref target="RFC7432"/>,
   				section 7.2) with the MUST requirement of the Ethernet Tag ID being
   				set to zero.
   			</t>
      
      		<t>
	      		Differently, the VLAN-Aware Bundle Service Interface follows a
   				bundling of multiple broadcast domains, with each having its own
   				bridge table, into a single EVI. This refers to the definition of
   				Multiple Broadcast Domain per EVI as described in <xref target="RFC8365"/> section
   				5.1.2 Option 2. The advantage of this model allows a single RD/RT per
   				broadcast domain, which becomes less relevant when VLAN-Based uses auto-
   				derivation of RD/RT. With VLAN-Aware Bundle Service, RT Constraint,
   				as defined in <xref target="RFC4684"/>, does not help to reduce the dissemination of
   				routes for a BD to the PEs attached to that BD. This is given by the
   				nature of the bundle service where the RT is not sufficient to
   				identify the MAC-VRF and corresponding bridge table. The differences
   				between the two modes of Service Interfaces, namely VLAN-Based and
   				VLAN-Aware Bundle Service Interface, is in the definition of the
   				Ethernet Tag field within the EVPN routes. While VLAN-Based Service
   				Interface defines the EtherTag must be set to zero, the VLAN-
   				Aware Bundle Service Interface uses the VID within the EtherTag to
   				identify the bridge table within the MAC-VRF. These two requirements
   				are orthogonal and as a result make the interoperability of the two
   				types mutually exclusive, interoperability is not achievable
   				(Figure 1).
   			</t>
	
			<t>
			<figure>
				<name>Interoperability of Service Interface Types</name>
				<artwork><![CDATA[
 VLAN-Aware Bundle                             VLAN-Based
 Service Interface                             Service Interface
+--------------------------+        +--------------------------+    
| PE1                      |        |                      PE2 |    
|  +---------+  +--------+ |        | +--------+  +---------+  |    
|  | +-----+ |  |        | |        | |        |  |         |  |   
+-----| BD2 | +--+        | |--------| |        +--+         |---+  
||  | +-----+ |  |        | |        | |        |  |         |  || 
||  |         |  |        | |        | |        |  |         |  || 
||  |MAC-VRF1 |  |IP-VRF1 | |        | |IP-VRF1 |  |MAC-VRF1 |  || 
||  +---------+  +--------+ |        | +--------+  +---------+  || 
||                          |        |                          || 
||                  +-----+ |        | +-----+                  || 
||                  | BGP | |        | | BGP |                  || 
||                  +-----+ |        | +-----+                  || 
|+--------------------------+        +--------------------------+|   
|            2|EthTag [2]| -----><----- 2|EthTag [0]|            |   
|                                                                |   
| +------+                                              +------+ | 
+-|M1/IP1!                                              |M2/IP2!-+   
  +------+                                              +------+
				   ]]>
			   </artwork>
			</figure>
     		</t>


      		<t>
	      		As illustrated in Figure 1, the MAC/IP routes exchanged by PE1 and
   				PE2 contain Ethernet Tags 2 and 0 respectively. The receiving PE will
   				not process these routes and will normally discard them (treat-as-
   				withdraw).
   			</t>
      
      		<t>
	      		By extending the requirements currently present, an interoperability
   				is achievable. The adjustment would be as follows.
   			</t>
   		
			<section title="VLAN-Aware Bundle Service PE" anchor="VlanBundPE">
	
				<t>
					In case of VLAN Aware Bundle Service Interface on the receiving PE
					and with the consideration of VLAN Based Service Interface on the
					advertising PE:</t>
				<ul>
					<li>For Service Interfaces: VLAN-Aware Bundle and VLAN-Based.</li>
					<li>For Integrated Routing and Bridging (IRB): Asymmetric IRB and Symmetric IRB.</li>
					<li>For IRB connectivity: interface-less and interface-ful unnumbered IRB.</li>
				
				<li> MUST operate in Single Broadcast Domain per EVI.</li>
				<li> Multiple Broadcast Domain per EVI case is not considered.</li>
				<li> Must allow to send and receive zero EtherTag.</li>
				<li> The import of routes is performed based on the import policy
					(route-target).</li>
				<li> 
					With single bridge table per MAC-VRF, additional evaluation of the
					EtherTag field is not required; the bridge table is sufficiently
					defined by the import route-target. Using a single MAC-VRF with per-
					BD route-target would deviate from the VLAN-Based Service Interface
					and would create a new interoperability permutation.</li>
				<li>  
					No Change to data-plane operation, the MPLS label identifies MAC-
					VRF + bridge-table, or the VNI identifies the MAC-VRF + the bridge-
					table.
				</li>
				</ul>
			</section>
			
			<section title="VLAN-Based Service PE" anchor="VlanBundPESer">
				<ul>
				  <li>Operates in Single Broadcast Domain per EVI.</li>
				</ul>
			  	<t>In case of VLAN Based Service Interface on the receiving PE and with
			   the consideration of VLAN Based Service Interface on the advertising
			   PE:</t>
			  	<ul>
			  	  <li>MUST allow receiving of non-zero EtherTag.</li>
			  	</ul>
				<t>- No Change in control-plane operation, the EVI import policy (route-
			   target) identifies the broadcast domain (bridge-table) within a MAC-
			   VRF.</t>
				<t>- No Change to data-plane operation, the MPLS label identifies MAC-
			   VRF + bridge-table, or the VNI identifies the MAC-VRF + the bridge-
			   table.</t>
				<t>While the expansion introduces additional configuration requirement
			   for the VLAN-Aware Bundle Service Interface, it also allows for
			   broader interoperability in the eventuality of vendor "A" only
			   implemented VLAN-Based while vendor "B" only implemented VLAN-Aware
			   Bundle Service Interface.</t>
		   </section>
	   </section>
	   
		<section title="Service Interface Interoperability Mode of Operation" anchor="ServiceInft">
			<t>
				When Service Interface interoperability is required, a given PE
			   	should follow this section's procedures for all its broadcast domains
			   	(BDs) and not just the BDs that need interoperability.</t>
			<t>For those BDs where interoperability between VLAN-Aware Bundle and
			   	VLAN-Based Service Interface is needed, ignoring the presence of the
			   	EVPN routes Ethernet Tag ID on the PEs supporting VLAN-Based mode is
			   	not enough. Each PE needs to clearly signal what mode it supports, so
			   	that all the PEs attached to the same EVI can understand in what mode
			   	the EVI operates.</t>
			<t>
				Consider a scenario where PE1 is attached to the BD range BD1-10 and
			   	it operates in VLAN-Aware mode, whereas PE2 is attached to the BD
			   	range BD7-20 and operates in VLAN-Based mode. Interoperability is
			   	required for the intersecting BDs, i.e., BD7-10.</t>
			<t>
				For PE1, this means BD7-10 need to be separated into a dedicated MAC-
			   	VRF each. EVPN routes for each of these four MAC-VRFs MUST be
			   	advertised by PE1 with an Ethernet Tag ID of zero. In this way, PE1
			   	indicates the use of VLAN-Based mode for those BDs. On reception, PE1
			   	imports the BD7-10 routes based on the Route Target and ignoring the
			   	Ethernet Tag ID, as the Route Target alone is sufficient to identify
			   	the correct MAC-VRF and Bridge Table. The remaining BDs on PE1 (range
			   	BD1-6) continue operating in VLAN-Aware Bundle mode.</t>
			<t>
				In the same example, other PEs attached to BD1-6 must still process
			   	the received Ethernet Tag ID in the EVPN routes from PE1, so that
			   	they can identify the correct Bridge Table in a given MAC-VRF.</t>
			<t>
				PE2 operates in VLAN-Based mode for BD7-20, as per <xref target="RFC7432"/> and
			   	<xref target="RFC8365"/>. PE2's EVPN route advertisements for BD7-20 will include
			   	individual Route Targets per BD and an Ethernet Tag ID of zero. On
			   	reception, PE2 identifies the MAC-VRF and Bridge Table solely based
			   	on the Route Target.</t>
    	</section>
	</section>

	<section title="Interoperability for different IRB Types" anchor="Interop">
		<section title="Asymmetric IRB and Symmetric IRB" anchor="AsymmetIRB1">
			<t>
				The differences in the two inter-subnet forwarding modes, namely
				Asymmetric IRB and Symmetric IRB, are beyond just the information
				difference in the control-plane from an EVPN Route Type 2 perspective
				(EVPN MAC/IP Advertisement route). The two IRB modes have significant
				differences in inter-subnet forwarding behavior and as a result
				different operation during label imposition or encapsulation.</t>
			<t>
				With the Asymmetric IRB mode, the ingress PE performs a "bridge-and-
				route" operation while the egress PE follows a "bridge-only"
				approach. Differently, the forwarding behavior in Symmetric IRB mode
				performs a "bridge-and-route" operation on the ingress PE followed by
				a "route-and bridge" operation at the egress PE. The significance in
				difference is not only in the forwarding behavior itself but also
				around how the respective EVPN attribute are used for driving the
				inter-subnet operation. More specifically, in the case of inter-
				subnet forwarding with Asymmetric IRB, next to the MAC and IP address
				present in the EVPN Route Type 2, only MPLS Label1 is used towards
				the egress PE to specify the MAC-VRF and respective Bridge-Domain for
				forwarding. In inter-subnet forwarding with Symmetric IRB, next to
				the MAC and IP address present in the EVPN Route Type 2, MPLS Label2</t>
			<t>
				associated with the IP-VRF is used for the inter-subnet forwarding
				operation towards egress PE. To facilitate interoperability between
				Asymmetric and Symmetric IRB, the usage of EVPN Route Type 2 with
				both MAC and IP address populated is considered. This is to not expand
				the requirement of additional capability exchange between PEs.
				Complementing Asymmetric IRBs absence of MPLS Label2 with an EVPN
				Route Type 5 (IP Prefix Advertisement) is not part of this procedure.</t>
			<t>
				The respective forwarding behaviors are described in <xref target="RFC9135"/>. The
				following steps are required to ensure the interoperability between
				the Asymmetric and Symmetric IRB modes.</t>
			
			<t>
				<figure>
					<name>Asymmetric IRB and Symmetric IRB</name>
					<artwork><![CDATA[
  Asymmetric IRB                                   Symmetric IRB 
+--------------------------+        +--------------------------+    
| PE1                      |        |                      PE2 |    
|  +---------+             |        |             +---------+  |    
|  | +-----+ |             |        |             | +-----+ |  |   
+-----| BD0 | +-IRB0-+      |        |     +-IRB0--+ | BD0 | |  |   
||  | +-----+ |      |      |        |     |       | +-----+ |  |  
||  |         |  +---+----+ |        | +---+----+  |         |  |    
||  |MAC-VRF1 |  |        | |        | |        |  |MAC-VRF1 |  |    
||  +---------+  |IP-VRF1 | |--------| |IP-VRF1 |  +---------+  |    
||               |        | |        | |        |               |   
||  +---------+  +---+----+ |        | +---+----+  +---------+  |    
||  | +-----+ |      |      |        |     |       | +-----+ |  |    
||  | | BD2 | +-IRB2-+      |        |     +-IRB2--+ | BD2 |-----+  
||  | +-----+ |             |        |             | +-----+ |  || 
||  |         |     +-----+ |        | +-----+     |         |  || 
||  |MAC-VRF2 |     | BGP | |        | | BGP |     |MAC-VRF2 |  || 
||  +---------+     +-----+ |        | +-----+     +---------+  || 
||                          |        |                          ||
|+--------------------------+        +--------------------------+|   
|       2|MAC/IP, 1 Label| -----><----- 2|MAC/IP, 2 Label|       |   
|                                                                |   
| +------+                                              +------+ | 
+-|M1/IP1!                                              |M2/IP2!-+   
  +------+                                              +------+
			  ]]>
					</artwork>
				</figure>
			</t>
			
			<t>
				Figure 2 illustrates the overview of an Asymmetric IRB PE (PE1) and
				a Symmetric IRB PE (PE2) within an interoperability deployment
				scenario. Attached to PE1, end-point M1/IP1 is attached to BD0 within
				MAC-VRF1. Respectively, on PE2 end-point M2/IP2 is connected via
				attachment circuit to BD2 positioned within MAC-VRF2. IRB0 and IRB2</t>
			<t>
				represent the host-facing IRB interface for inter-subnet
				communication between the end-points located in the different IP
				subnet. The IRB interfaces for a common MAC-VRF/BD on PE1 and PE2 use
				the same IP address. With the IRB modes on PE1 being Asymmetric IRB
				and PE2 being Symmetric IRB, the MPLS Label fields, as part of the
				MAC/IP routes exchanged between the PEs, are different. PE1's update
				contains a single label, representing MPLS Label1 used for bridging
				purposes. PE2's advertisement contains two labels, one for bridging
				and one for routing, as part of the MAC/IP route. While PE1 receives
				all information necessary from PE2, PE2 is missing information
				necessary for its routing operation. As a result, inter-subnet
				routing between PE1 and PE2 is not achieved.</t>
			<t>
				By following the current existing forwarding behavior as described in
				<xref target="RFC9135"/>, interoperability is theoretically achievable without
				changes in the control-plane format. Nevertheless, there are
				additional steps required that involves the local forwarding behavior
				of the PE with Symmetric IRB mode.</t>

			<section title="Asymmetric IRB PE" anchor="AsymmetIRB2">

				<t>
					In case of Asymmetric IRB as the advertising PE and with Symmetric
					IRB on the receiving PE:
				</t>
				<t>- 
					Asymmetric IRB PE MUST send MAC and IP information with MPLS Label1
					as described in <xref target="RFC9135"/>.
				</t>
				<t>
					In case of Symmetric IRB as the advertising PE and with Asymmetric
					IRB on the receiving PE:
				</t>
				<t>- 
					Asymmetric IRB PE MUST be able to ignore MPLS Label2; <xref target="RFC9135"/>
					already considers this.
				</t>
			</section>
			
			<section title="Symmetric IRB PE" anchor="AsymmetIRB3">
				<t>In case of Symmetric IRB as the advertising PE and with Asymmetric
				IRB on the receiving PE:</t>
				<ul>
				  <li>Symmetric IRB PE has no additional requirements.</li>
				</ul>
				<t>In case of Asymmetric IRB as the advertising PE and with Symmetric
				IRB on the receiving PE:</t>
				<t>- Symmetric IRB PE requires to add the host-binding information, MAC
				and IP, and associates them to the adjacency (ARP/ND) table facing
				the PE with Asymmetric IRB; this is in addition of adding the MAC
				address into the MAC-VRF table, using MPLS Label1. Since there is no
				MPLS Label2 or Route-Target for the IP-VRF, the Host IP is not</t>
				<t>specifically added to IP-VRF table.</t>
				<t>- Symmetric IRB PE must have defined all BD, MAC-VRF and IRB
				interfaces of the Asymmetric IRB PE.</t>
			</section>
		</section>
		
		<section title="IRB Interop Mode of Operation" anchor="IRBInteropMode1">
			<t>Interoperability between the Asymmetric IRB and Symmetric IRB mode
			follows specific defined behavior that is predominantly required on
			the PE that operates in the Symmetric IRB mode. Nevertheless, in
			support for the interoperability, the PE operating in Asymmetric IRB
			must accommodate the following two minimum requirements (with
			references to Figure 2):</t>
			<t>1) The PE that operates in Asymmetric IRB mode (PE1), MUST send the
			MAC/IP route including the Host IP address and only MPLS Label1.</t>
			<t>2) The PE with Asymmetric IRB (PE1) must accept the MAC/IP routes
			sent from PE2 (Symmetric IRB), while ignoring the additional
			information of MPLS Label2 and Route-Target of the IP-VRF.</t>
			<t>In reference to 1), the PE MUST always send the end-point MAC
			address, Host IP address and related MPLS Label1 as part of the
			MAC/IP route towards the PE with Symmetric IRB (PE2). This route will
			be sent only with MPLS Label1 and the Route-Target of the matching
			MAC-VRF to achieve bridging. In reference to the illustration in
			Figure 2, PE1 must generate and advertise an EVPN MAC/IP route using:</t>
				<ul>
				  <li>MAC Length of 48</li>
				  <li>MAC Address of M1</li>
				  <li>IP Length of 32 / 128</li>
				  <li>IP Address of IP1</li>
				  <li>Label for MAC-VRF1</li>
				  <li>Route-Target of MAC-VRF1</li>
				  <li>Next-Hop PE1</li>
				</ul>
			<t>For completeness of the requirements and in reference of 2), the
			MAC/IP route advertised from the PE operating in Symmetric IRB (PE2)
			is as follow:</t>
				<ul>
				  <li>MAC Length of 48</li>
				  <li>MAC Address of M2</li>
				  <li>IP Length of 32 /128</li>
				  <li>IP Address of IP2</li>
				  <li>Label for MAC-VRF2, IP-VRF1</li>
				  <li>Route-Target of MAC-VRF2, IP-VRF1</li>
				  <li>Next-Hop PE2</li>
				</ul>
			<t>As defined in 2), the Label and Route-Target information for IP-VRF1
			MUST be ignored by PE1 (PE with Asymmetric IRB).</t>
			<t>With PE2 operating in Symmetric IRB and with enabled interop mode,
			the MAC/IP route from PE1 (Asymmetric IRB) is processed in the
			respective bridging, routing and adjacency table. Based on the Route-
			Target for MAC-VRF1, the MAC address M1 will be imported into MAC-
			VRF1 respectively and placed within BD0. In addition, the host-
			binding information M1/IP1 MUST be installed within PE2's adjacency
			table. Subsequently, on PE2 the MAC address M1 and the host-binding
			information (adjacency table entry) of M1/IP1 MUST point towards PE1
			as the next-hop. With no presence of the Route-Target for IP-VRF1,
			the IP address IP1 will not be specifically imported into IP-VRF1 and
			is not associated with a MPLS Label2. As a result of the
			interoperability, the additional efficiency provided by Symmetric IRB
			with regards to preserving adjacency table exhaustion is reduced; this
			is specifically when communicating with an Asymmetric IRB based
			egress PE. In contrast, the interop mode allows for communication
			between the different IRB modes. As a result, in the eventuality that
			vendor "A" only provides Asymmetric IRB, while vendor "B" only has
			Symmetric IRB available, interoperability for inter-subnet forwarding
			can be seamlessly achieved. In addition, two further benefits are
			present by implementing an Asymmetric/Symmetric IRB Co-Existence on the
			same PE (dual-mode PE).</t>
			<t>- A dual-mode PE can seamlessly communicate with PEs that are either
			in Asymmetric or in Symmetric IRB mode.</t>
			<t>- A dual-mode PE can act as Anchor for interconnecting Symmetric IRB
			and Asymmetric IRB based PE's (deployment restrictions might apply).</t>
		</section>
	</section>


	<section title="Interoperability for different IRB Core Connectivity Modes" anchor="InteropCore">
		<section title="Interface-Less and Interface-Ful Unnumbered IRB" anchor="Intfless0">
			<t>
				The two modes, namely interface-less and interface-ful Unnumbered SBD
				IRB, are closely related with regards to the information required in
				the EVPN Route Type 5. While interface-less provides all information
				for the IP prefix advertisement within the EVPN Route Type 5, in the
				case of interface-ful Unnumbered SBD IRB, an additional EVPN Route
				Type 2 is required for the next-hop recursive lookup. From a
				forwarding behavior, both approaches are similar and follow a
				symmetric routing approach but are not interoperable. Note as per
				<xref target="RFC9136"/> the interface-ful Unnumbered SBD IRB is an OPTIONAL mode.
			</t>
			<t>
				<figure>
					<name>
						Interoperability of different IRB Core Connectivity Mode (unnumbered)
					</name>
					<artwork><![CDATA[
Interface-Less                        Interface-Ful Unnumbered IRB 
+--------------------------+        +--------------------------+    
| PE1                      |        |                      PE2 |    
|               +--------+ |        | +--------+               |    
|               |        | |        | |        |               |  
+----------------+        | |--------| |        +----------------+  
||               |        | |        | |        |               || 
||               |        | |        | |        |               || 
||               |IP-VRF1 | |        | |IP-VRF1 |               || 
||               +--------+ |        | +--------+               || 
||                          |        |                          || 
||                  +-----+ |        | +-----+                  || 
||                  | BGP | |        | | BGP |                  || 
||                  +-----+ |        | +-----+                  || 
|+--------------------------+        +--------------------------+|   
|                  2|None| ----->                                |   
|                                <----- 5|No Label|              |   
|                                                                |   
| +-------+                                            +-------+ | 
+-|TS1/SN1|                                            |TS2/SN2!-+   
  +-------+                                            +-------+
					  ]]>
					</artwork>
				</figure>
			</t>
			<t>
				The illustration in Figure 3 represents the possible deployment
				scenario between two different Core IRB Connectivity modes.
				Specifically, PE1 is operating with interface-less Core IRB Mode
				while PE2 operates with the interface-ful Unnumbered SDB IRB mode;
				both operate without interoperability capabilities. Attached to PE1
				and PE2 respectively, Tenant System 1 (TS1) and Tenant System 2 (TS2)
				with different IP subnets are present. TS1 attached on PE1 as well as
				TS2 attached to PE2 are represented in a common IP-VRF (IP-VRF1),
				sharing a common Route-Target between the PEs. With the different IRB
				Core Connectivity modes on PE1 and PE2 respectively, the differences
				in IP prefix advertisements as described in <xref target="RFC9136"/> are present.
				PE1 advertises only a single EVPN Route Type 5 (IP Prefix Route) for
				TS1 using the fields defined for the interface-less mode:
			</t>
			<t>EVPN Route Type 5:</t>
				<ul>
				  <li>IP Length of 0 to 32 / 0 to 128</li>
				  <li>IP Address of SN1</li>
				  <li>Label for IP-VRF1</li>
				  <li>GW IP Address set to zero</li>
				  <li>Route-Target of IP-VRF1</li>
				  <li>Router's MAC Extended Community of PE1</li>
				  <li>Next-Hop PE1</li>
				</ul>
			<t>
				Differently, PE2 advertises an EVPN Route Type 2 (MAC/IP Route) next
				to the EVPN Route Type 5 (IP Prefix Route). The MAC/IP Route supports
				the requirement for recursive next-hop resolution for the next-hop
				used in the IP Prefix Route. Below the fields used in the Route Type
				5 and respective Route Type 2 according to the interface-ful
				Unnumbered IRB mode:
			</t>
			<t>EVPN Route Type 5:</t>
				<ul>
				  <li>IP Length of 0 to 32 / 0 to 128</li>
				  <li>IP Address of SN1</li>
				  <li>Label SHOULD be set to 0</li>
				  <li>GW IP Address SHOULD be set to zero</li>
				  <li>Route-Target of IP-VRF1</li>
				  <li>Router's MAC Extended Community of PE2</li>
				  <li>Next-Hop PE2</li>
				</ul>
			<t>EVPN Route Type 2:</t>
				<ul>
				  <li>MAC Length of 48</li>
				  <li>MAC Address of PE2</li>
				  <li>IP Length of 0</li>
				  <li>Label for IP-VRF1</li>
				  <li>Route-Target of IP-VRF1</li>
				  <li>Next-Hop PE2</li>
				</ul>
			<t>
				While PE1 is missing the MPLS Label for the IP-VRF from PE2, PE2 is
				missing the MPLS Label information and the necessary info for the
				next-hop recursion. As a result, Routing with IP Prefix Advertisement
				between PE1 and PE2 is not achieved.
			</t>
			<t>
				By advertising an additional EVPN Route Type 2 from interface-less
				(PE1) and by advertising the MPLS Label as part of EVPN Route Type 5
				from PE2, interoperability is achievable. The specific mode of
				operation would be as per the following two sections and refers to
				Figure 3 and Figure 4.
			</t>
			
			<section title="Interface-Less PE" anchor="Intfless1">
	
				<t>
					In case of interface-less on the advertising PE and with the
					consideration of interface-ful Unnumbered IRB as the receiving PE:</t>
				<t>
					The interface-less PE MUST generate and advertise an EVPN Route Type 2
				</t>
					<ul>
					  <li>MAC Length of 48</li>
					  <li>MAC Address with "Router MAC"</li>
					  <li>IP Length of 0</li>
					  <li>Label for IP-VRF</li>
					  <li>Route-Target of IP-VRF</li>
					</ul>
				<t>
					In case of interface-less on the receiving PE and with the 
					consideration of interface-ful Unnumbered IRB as the advertising PE:</t>
				<t>- 
					MUST ignore EVPN Route Type 2 with MPLS Label and route-target
					matching the IP-VRF because there is no MAC-VRF defined matching
					this information.</t>
			</section>
		
			<section title="Interface-Ful Unnumbered IRB" anchor="Intfless2">

				<t>In case of interface-ful Unnumbered on the advertising PE and with
				the consideration of interface-less as the receiving PE:</t>
				<t>- Shall advertise MPLS Label for IP-VRF in EVPN Route Type 5 with
				matching route-target.</t>
				<t>In case of interface-ful Unnumbered on the receiving PE and with the
				consideration of interface-less as the advertising PE:</t>
				<ul>
				  <li>No Additions Required.</li>
				</ul>
				<t>
					<figure>
						<name>
						Interop of different IRB Core Connectivity Types (unnumbered)
						</name>
						<artwork><![CDATA[
														 
Interface-Less                       Interface-Ful Unnumbered IRB 
+--------------------------+        +--------------------------+    
| PE1                      |        |                      PE2 |    
|               +--------+ |        | +--------+               |    
|               |        | |        | |        |               |  
+----------------+        | |--------| |        +----------------+  
||               |        | |        | |        |               || 
||               |        | |        | |        |               || 
||               | IP-VRF | |        | | IP-VRF |               || 
||               +--------+ |        | +--------+               || 
||                          |        |                          || 
||                  +-----+ |        | +-----+                  || 
||                  | BGP | |        | | BGP |                  || 
||                  +-----+ |        | +-----+                  || 
|+--------------------------+        +--------------------------+|   
|              2|RMAC/RIP| ----->                                |   
|                                <----- 5|Label|                 |   
|                                                                |   
| +-------+                                            +-------+ | 
+-|TS1/SN1|                                            |TS2/SN2!-+   
  +-------+                                            +-------+
						 ]]>
						</artwork>
					</figure>
				</t>

				<t>
					Illustrated in Figure 4 are the additional requirements for
					interface-less IRB Core Connectivity mode, specifically the MAC/IP
					Route (EVPN Route Type 2) necessary for PE2's next-hop recursion.
					Also, the MPLS Label addition within PE2's IP Prefix route (EVPN Route
					Type 5) is represented, which is required for advertisement by interface-ful 
					Unnumbered IRB towards an interface-less PE (PE1)</t>
				<t>
					The interop mode introduces additional control-plane advertisements
					from an Interface-less perspective. This is necessary to allow
					interface-ful Unnumbered SBD IRB to perform the recursive lookup
					required. From a EVPN Type 5 perspective between the two types, most
					of the fields are already equally defined and populated as per
					<xref target="RFC9136"/>. Exception is the IP-VRF Label, which is required to be
					added in the interface-ful Unnumbered SBD IRB's EVPN Type 5. In
					addition, the Interface-less addition allows the Co-Existence of both
					types on the same PE (dual-mode PE). Such a dual-mode PE can
					communicate at the same time with PEs that are in Interface-less or
					in interface-ful Unnumbered SBD IRB mode.
				</t>
				<t>
					The disadvantage of the additional advertisement has to be put into
					relation to advantage of successful interoperability where eventually
					vendor "A" only implemented interface-less while vendor "B" only
					implemented interface-ful Unnumbered SBD IRB.
				</t>
			</section>
		</section>
		
			<section title="Tunnel Encapsulation Consideration" anchor="Tunnel1">

			<t>
				With regards to IRB core connectivity both solutions, namely interface-
				less and interface-ful, provide a solution for Layer 3 connectivity
				among the IP-VRFs. Even as the functional result of both modes is the
				same, there are important considerations with regards to tunnel
				encapsulations.
			</t>
			<t>
				<xref target="RFC9135"/> section 4 considers the choice for the NVO tunnel should be
				dictated by the tunnel capabilities. For example for the IP-VRF-to-
				IP-VRF model with interface-less, the NVO tunnel for MPLS needs to be
				IP NVO and for VXLAN needs to be Ethernet NVO.
			</t>
			<t>
				With the "IP-VRF-to-IP-VRF" model that is used in interface-ful
				(numbered or unnumbered), section 4.4.2 or 4.4.3 respectively
				describe the solution to accommodate Ethernet NVO tunnels (VXLAN or
				GPE, GENEVE, MPLS with MAC payload) only. In the case of interface-
				ful unnumbered, the Router-MAC Extended Community is always signaled
				via EVPN update message, which implies the presence of a MAC payload.
				IP NVO tunnels are not applicable to these two use-cases/models</t>
			<t>
				Depending on the use of NVO tunnels, interoperability between
				interface-less and interface-ful unnumbered requires additional
				changes on the Tunnel Encapsulation mode. This Internet-Draft
				considers the usage of a compatible NVO Tunnel mode between a PE
				operating in interface-less and a PE operating interface-ful unnumbered
				mode.
			</t>
		</section>
	</section>

	<section title="Security Considerations" anchor="Sec">
		<t>
			The security considerations of <xref target="RFC7432"/>, <xref target="RFC8365"/>, and 
			<xref target="RFC9136"/> apply to this document.
			This document defines procedures that rely on existing BGP control-plane
			and data-plane mechanisms and therefore introduce no new security
			considerations beyond those described in the referenced documents.
		</t>
	</section>
	
	<section title="IANA Considerations" anchor="Iana">
		<t>
			This document has no IANA actions.
		</t>
	</section>
	
	<section title="Conclusion" anchor="Concl">
		<t>
			With minimal additions, the most common EVPN types for Virtual
		Identifiers to EVI Mapping, Integrated Routing and Bridging and IP
		Prefix Advertisement can be made interoperable. The aim for
		interoperability doesn't remove the requirement for optimized types
		for different use-cases but allows flexibility and basic
		interoperability.
		</t>
		
		<section title="Demonstration of Applicability" anchor="Demonstration">
			<t>
				Cisco, Juniper and Nokia demonstrated successfully the ability of
				EVPN interoperability modes during EANTC's yearly "Multi-Vendor
				Interoperability Test". The Whitepaper can be obtained through EANTC
				with the latest version being available at <xref target="EANTC"/>.
			</t>
			
			<section title="Service Interface Interoperability" anchor="ServiceType1">
				<t>
					A proof of the benefit with this interoperability mode has already
					been demonstrated during EVPN Multi-Vendor interoperability testing
					and also, in production environments. Specifically, Cisco and Nokia's
					VLAN-Based Service Interface successfully demonstrated interoperability with
					Juniper's VLAN-Aware Bundle Service Interface.</t>
			</section>
		
			<section title="IRB Types" anchor="IRBTypes1">
				<t>
					A proof of the benefit with this interoperability mode has already
					successfully demonstrated during EVPN Multi-Vendor interoperability
					testing. Specifically, Cisco operated in a Hybrid IRB (Dual-Mode)
					mode while other vendors operated in an Asymmetric IRB mode.</t>
				<t>
					Forwarding was achieved through dynamic detection of the alternate
					vendor PE's mode and adjustment to Asymmetric IRB for these specific
					BDs. Communication for all other BDs continued to be Symmetric IRB.
				</t>
			</section>
			
			<section title="IRB Core Connectivity Types" anchor="IRBConTypes1">
				<t>
					A proof of an interoperability mode between interface-less and
					interface-ful Unnumbered SBD IRB has already been demonstrated in
					production environments and during EVPN Multi-Vendor interoperability
					testing. Specifically, Cisco's addition for Interface-less is
					successfully deployed with Nokia's and Nuage's interface-ful
					Unnumbered SBD IRB at customers
				</t>
			</section>
			
		</section>
	</section>

</middle>
  
<back>
	    
	<!-- References split into informative and normative -->
		<references title="Normative References">
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.2119.xml"/>
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.7432.xml"/>
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.8174.xml"/>
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.8365.xml"/>
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.9135.xml"/>
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.9136.xml"/>
		</references>
		
		
		<references title="Informative References">

			<!-- <xi:include
			href="https://www.rfc-editor.org/refs/bibxml6/reference.IEEE.802.1Q_2018.xml"/> -->
			<xi:include href="https://www.rfc-editor.org/refs/bibxml/reference.RFC.4684.xml"/>
			<reference anchor="EANTC" target="https://eantc.de/wp-content/uploads/2023/04/EANTC-MPLSSDNNFV2019-WhitePaper-v1.3.pdf">
				<front>
					<title>Multi-Vendor Interoperability Test</title>
					<author>
						<organization>EANTC</organization>
					</author>
					<date year="February 2019"/>
				</front>
			</reference>
		</references>
		
		<section title="Contributors" anchor="Contributors1" numbered="false" toc="include">
		<t>In addition to the authors listed, the following individuals also contributed to this document:</t>
		<author fullname="Krishnaswamy Ananthamurthy" initials="K." surname="Ananthamurthy">
		<organization>Cisco</organization>
		<address>
		  <postal>
			<street>170 W. Tasman Drive</street>
			<street/>
			<city>San Jose</city>
			<code>95134</code>
			<region>CA</region>
			<country>USA</country>
		  </postal>
		  <email>kriswamy@cisco.com</email>
		</address>
		</author>
	</section>

</back>
</rfc>