RFC 9161




Internet Engineering Task Force (IETF)                   J. Rabadan, Ed.
Request for Comments: 9161                                  S. Sathappan
Updates: 7432                                                 K. Nagaraj
Category: Standards Track                                     G. Hankins
ISSN: 2070-1721                                                    Nokia
                                                                 T. King
                                                                  DE-CIX
                                                            January 2022


Operational Aspects of Proxy ARP/ND in Ethernet Virtual Private Networks

Abstract



   This document describes the Ethernet Virtual Private Network (EVPN)
   Proxy ARP/ND function augmented by the capability of the ARP/ND
   Extended Community.  From that perspective, this document updates the
   EVPN specification to provide more comprehensive documentation of the
   operation of the Proxy ARP/ND function.  The EVPN Proxy ARP/ND
   function and the ARP/ND Extended Community help operators of Internet
   Exchange Points, Data Centers, and other networks deal with IPv4 and
   IPv6 address resolution issues associated with large Broadcast
   Domains by reducing and even suppressing the flooding produced by
   address resolution in the EVPN network.

Status of This Memo



   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc9161.

Copyright Notice



   Copyright (c) 2022 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
     1.1.  The Data Center Use Case
     1.2.  The Internet Exchange Point Use Case
   2.  Terminology
   3.  Solution Description
     3.1.  Proxy ARP/ND Sub-functions
     3.2.  Learning Sub-function
       3.2.1.  Proxy ND and the NA Flags
     3.3.  Reply Sub-function
     3.4.  Unicast-Forward Sub-function
     3.5.  Maintenance Sub-function
     3.6.  Flood (to Remote PEs) Handling
     3.7.  Duplicate IP Detection
   4.  Solution Benefits
   5.  Deployment Scenarios
     5.1.  All Dynamic Learning
     5.2.  Dynamic Learning with Proxy ARP/ND
     5.3.  Hybrid Dynamic Learning and Static Provisioning with Proxy
           ARP/ND
     5.4.  All Static Provisioning with Proxy ARP/ND
     5.5.  Example of Deployment in Internet Exchange Points
     5.6.  Example of Deployment in Data Centers
   6.  Security Considerations
   7.  IANA Considerations
   8.  References
     8.1.  Normative References
     8.2.  Informative References
   Acknowledgments
   Contributors

   Authors' Addresses



1.  Introduction



   As specified in [RFC7432], the IP Address field in the Ethernet
   Virtual Private Network (EVPN) Media Access Control (MAC) / IP
   Advertisement route may optionally carry one of the IP addresses
   associated with the MAC address.  A Provider Edge (PE) may learn
   local IP->MAC pairs and advertise them in EVPN MAC/IP Advertisement
   routes.  Remote PEs importing those routes in the same Broadcast
   Domain (BD) may add those IP->MAC pairs to their Proxy ARP/ND tables
   and reply to local ARP Requests or Neighbor Solicitations (or
   "unicast-forward" those packets to the owner MAC), reducing and even
   suppressing, in some cases, the flooding in the EVPN network.

   EVPN and its associated Proxy ARP/ND function are extremely useful in
   Data Centers (DCs) or Internet Exchange Points (IXPs) with large
   Broadcast Domains, where the amount of ARP/ND flooded traffic causes
   issues on connected routers and Customer Edges (CEs).  [RFC6820]
   describes the address resolution problems in large DC networks.

   This document describes the Proxy ARP/ND function in [RFC7432]
   networks, augmented by the capability of the ARP/ND Extended
   Community [RFC9047].  From that perspective, this document updates
   [RFC7432].

   Proxy ARP/ND may be implemented to help IXPs, DCs, and other
   operators deal with the issues derived from address resolution in
   large Broadcast Domains.

1.1.  The Data Center Use Case



   As described in [RFC6820], the IPv4 and IPv6 address resolution can
   create a lot of issues in large DCs.  In particular, the issues
   created by IPv4 Address Resolution Protocol procedures may be
   significant.

   On one hand, ARP Requests use broadcast MAC addresses; therefore, any
   Tenant System in a large Broadcast Domain will see a large amount of
   ARP traffic, which is not addressed to most of the receivers.

   On the other hand, the flooding issue becomes even worse if some
   Tenant Systems disappear from the Broadcast Domain, since some
   implementations will persistently retry sending ARP Requests.  As
   [RFC6820] states, there are no clear requirements for retransmitting
   ARP Requests in the absence of replies; hence, an implementation may
   choose to keep retrying endlessly even if there are no replies.

   The amount of flooding that address resolution creates can be
   mitigated by the use of EVPN and its Proxy ARP/ND function.

1.2.  The Internet Exchange Point Use Case



   The implementation described in this document is especially useful in
   IXP networks.

   A typical IXP provides access to a large Layer 2 Broadcast Domain for
   peering purposes (referred to as "the peering network"), where
   (hundreds of) Internet routers are connected.  We refer to these
   Internet routers as CE devices in this section.  Because of the
   requirement to connect all routers to a single Layer 2 network, the
   peering networks use IPv4 addresses in length ranges from /21 to /24
   (and even bigger for IPv6), which can create very large Broadcast
   Domains.  This peering network is transparent to the CEs and
   therefore floods any ARP Requests or NS messages to all the CEs in
   the network.  Gratuitous ARP and NA messages are flooded to all the
   CEs too.

   In these IXP networks, most of the CEs are typically peering routers
   and roughly all the Broadcast, Unknown Unicast, and Multicast (BUM)
   traffic is originated by the ARP and ND address resolution
   procedures.  This ARP/ND BUM traffic causes significant data volumes
   that reach every single router in the peering network.  Since the
   ARP/ND messages are processed in "slow path" software processors and
   they take high priority in the routers, heavy loads of ARP/ND traffic
   can cause some routers to run out of resources.  CEs disappearing
   from the network may cause address resolution explosions that can
   make a router with limited processing power fail to keep BGP sessions
   running.

   The issue might be better in IPv6 routers if Multicast Listener
   Discovery (MLD) snooping was enabled, since ND uses an SN-multicast
   address in NS messages; however, ARP uses broadcast and has to be
   processed by all the routers in the network.  Some routers may also
   be configured to broadcast periodic Gratuitous ARPs (GARPs)
   [RFC5227].  For IPv6, the fact that IPv6 CEs have more than one IPv6
   address contributes to the growth of ND flooding in the network.  The
   amount of ARP/ND flooded traffic grows linearly with the number of
   IXP participants; therefore, the issue can only grow worse as new CEs
   are added.

   In order to deal with this issue, IXPs have developed certain
   solutions over the past years.  While these solutions may mitigate
   the issues of address resolution in large Broadcast Domains, EVPN
   provides new more efficient possibilities to IXPs.  EVPN and its
   Proxy ARP/ND function may help solve the issue in a distributed and
   scalable way, fully integrated with the PE network.

2.  Terminology



   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 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

   ARP:        Address Resolution Protocol

   AS-MAC:     Anti-spoofing MAC.  It is a special MAC configured on all
               the PEs attached to the same BD and used for the
               duplicate IP detection procedures.

   BD:         Broadcast Domain

   BUM:        Broadcast, Unknown Unicast, and Multicast Layer 2 traffic

   CE:         Customer Edge router

   DAD:        Duplicate Address Detection, as per [RFC4861]

   DC:         Data Center

   EVI:        EVPN Instance

   EVPN:       Ethernet Virtual Private Network, as per [RFC7432]

   GARP:       Gratuitous ARP

   IP->MAC:    An IP address associated to a MAC address.  IP->MAC
               entries are programmed in Proxy ARP/ND tables and may be
               of three different types: dynamic, static, or EVPN-
               learned.

   IXP:        Internet Exchange Point

   IXP-LAN:    The IXP's large Broadcast Domain to where Internet
               routers are connected.

   LAG:        Link Aggregation Group

   MAC or IP DA:  MAC or IP Destination Address

   MAC or IP SA:  MAC or IP Source Address

   ND:         Neighbor Discovery

   NS:         Neighbor Solicitation

   NA:         Neighbor Advertisement

   NUD:        Neighbor Unreachability Detection, as per [RFC4861]

   O Flag:     Override Flag in NA messages, as per [RFC4861]

   PE:         Provider Edge router

   R Flag:     Router Flag in NA messages, as per [RFC4861]

   RT2:        EVPN Route type 2 or EVPN MAC/IP Advertisement route, as
               per [RFC7432]

   S Flag:     Solicited Flag in NA messages, as per [RFC4861]

   SN-multicast address:  Solicited-Node IPv6 multicast address used by
               NS messages

   TLLA:       Target Link Layer Address, as per [RFC4861]

   VPLS:       Virtual Private LAN Service

   This document assumes familiarity with the terminology used in
   [RFC7432].

3.  Solution Description



   Figure 1 illustrates an example EVPN network where the Proxy ARP/ND
   function is enabled.

                                                         BD1
                                                     Proxy ARP/ND
                                                    +------------+
   IP1/M1          +----------------------------+   |IP1->M1 EVPN|
    GARP --->Proxy ARP/ND                       |   |IP2->M2 EVPN|
   +---+      +--------+   RT2(IP1/M1)          |   |IP3->M3 sta |
   |CE1+------|   BD1  |    ------>      +------+---|IP4->M4 dyn |
   +---+      +--------+                 |          +------------+
                  PE1                    | +--------+ Who has IP1?
                   |           EVPN      | |   BD1  | <-----  +---+
                   |           EVI1      | |        | ----->  |CE3|
   IP2/M2          |                     | |        | IP1->M1 +---+
    GARP  --->Proxy ARP/ND               | +--------+   |    IP3/M3
     +---+    +--------+   RT2(IP2/M2)   |              |
     |CE2+----|   BD1  |    ------>      +--------------+
     +---+    +--------+                       PE3|    +---+
                  PE2                           | +----+CE4|
                   +----------------------------+      +---+
                                                 <---IP4/M4 GARP

                   Figure 1: Proxy ARP/ND Network Example

   When the Proxy ARP/ND function is enabled in a BD (Broadcast Domain)
   of the EVPN PEs, each PE creates a Proxy table specific to that BD
   that can contain three types of Proxy ARP/ND entries:

   Dynamic entries:
      Learned by snooping a CE's ARP and ND messages; for instance, see
      IP4->M4 in Figure 1.

   Static entries:
      Provisioned on the PE by the management system; for instance, see
      IP3->M3 in Figure 1.

   EVPN-learned entries:
      Learned from the IP/MAC information encoded in the received RT2's
      coming from remote PEs; for instance, see IP1->M1 and IP2->M2 in
      Figure 1.

   As a high-level example, the operation of the EVPN Proxy ARP/ND
   function in the network of Figure 1 is described below.  In this
   example, we assume IP1, IP2, and IP3 are IPv4 addresses:

   1.  Proxy ARP/ND is enabled in BD1 of PE1, PE2, and PE3.

   2.  The PEs start adding dynamic, static, and EVPN-learned entries to
       their Proxy tables:

       a.  PE3 adds IP1->M1 and IP2->M2 based on the EVPN routes
           received from PE1 and PE2.  Those entries were previously
           learned as dynamic entries in PE1 and PE2, respectively, and
           advertised in BGP EVPN.

       b.  PE3 adds IP4->M4 as dynamic.  This entry is learned by
           snooping the corresponding ARP messages sent by CE4.

       c.  An operator also provisions the static entry IP3->M3.

   3.  When CE3 sends an ARP Request asking for the MAC address of IP1,
       PE3 will:

       a.  Intercept the ARP Request and perform a Proxy ARP lookup for
           IP1.

       b.  If the lookup is successful (as in Figure 1), PE3 will send
           an ARP Reply with IP1->M1.  The ARP Request will not be
           flooded to the EVPN network or any other local CEs.

       c.  If the lookup is not successful, PE3 will flood the ARP
           Request in the EVPN network and the other local CEs.

   In the same example, if we assume IP1, IP2, IP3, and IP4 are now IPv6
   addresses and Proxy ARP/ND is enabled in BD1:

   1.  PEs will start adding entries in a similar way as they would for
       IPv4; however, there are some differences:

       a.  IP1->M1 and IP2->M2 are learned as dynamic entries in PE1 and
           PE2, respectively, by snooping NA messages and not by
           snooping NS messages.  In the IPv4 case, any ARP frame can be
           snooped to learn the dynamic Proxy ARP entry.  When learning
           the dynamic entries, the R and O Flags contained in the
           snooped NA messages will be added to the Proxy ND entries
           too.

       b.  PE1 and PE2 will advertise those entries in EVPN MAC/IP
           Advertisement routes, including the corresponding learned R
           and O Flags in the ARP/ND Extended Community.

       c.  PE3 also adds IP4->M4 as dynamic after snooping an NA message
           sent by CE4.

   2.  When CE3 sends an NS message asking for the MAC address of IP1,
       PE3 behaves as in the IPv4 example, by intercepting the NS,
       performing a lookup on the IP, and replying with an NA if the
       lookup is successful.  If it is successful, the NS is not flooded
       to the EVPN PEs or any other local CEs.

   3.  If the lookup is not successful, PE3 will flood the NS to remote
       EVPN PEs attached to the same BD and the other local CEs as in
       the IPv4 case.

   As PE3 learns more and more host entries in the Proxy ARP/ND table,
   the flooding of ARP Request messages among PEs is reduced and in some
   cases, it can even be suppressed.  In a network where most of the
   participant CEs are not moving between PEs and are advertising their
   presence with GARPs or unsolicited-NA messages, the ARP/ND flooding
   among PEs, as well as the unknown unicast flooding, can practically
   be suppressed.  In an EVPN-based IXP network, where all the entries
   are static, the ARP/ND flooding among PEs is in fact totally
   suppressed.

   In a network where CEs move between PEs, the Proxy ARP/ND function
   relies on the CE signaling its new location via GARP or unsolicited-
   NA messages so that tables are immediately updated.  If a CE moves
   "silently", that is, without issuing any GARP or NA message upon
   getting attached to the destination PE, the mechanisms described in
   Section 3.5 make sure that the Proxy ARP/ND tables are eventually
   updated.

3.1.  Proxy ARP/ND Sub-functions



   The Proxy ARP/ND function can be structured in six sub-functions or
   procedures:

   1.  Learning sub-function

   2.  Reply sub-function

   3.  Unicast-forward sub-function

   4.  Maintenance sub-function



   5.  Flood handling sub-function



   6.  Duplicate IP detection sub-function



   A Proxy ARP/ND implementation MUST at least support the Learning,
   Reply, Maintenance, and duplicate IP detection sub-functions.  The
   following sections describe each individual sub-function.

3.2.  Learning Sub-function



   A Proxy ARP/ND implementation in an EVPN BD MUST support dynamic and
   EVPN-learned entries and SHOULD support static entries.

   Static entries are provisioned from the management plane.  A static
   entry is configured on the PE attached to the host using the IP
   address in that entry.  The provisioned static IP->MAC entry MUST be
   advertised in EVPN with an ARP/ND Extended Community where the
   Immutable ARP/ND Binding Flag (I) is set to 1, as per [RFC9047].
   When the I Flag in the ARP/ND Extended Community is 1, the
   advertising PE indicates that the IP address must not be associated
   to a MAC other than the one included in the EVPN MAC/IP Advertisement
   route.  The advertisement of I = 1 in the ARP/ND Extended Community
   is compatible with any value of the Sticky bit (S) or sequence number
   in the [RFC7432] MAC Mobility Extended Community.  Note that the I
   bit in the ARP/ND Extended Community refers to the immutable
   configured association between the IP and the MAC address in the
   IP->MAC binding, whereas the S bit in the MAC Mobility Extended
   Community refers to the fact that the advertised MAC address is not
   subject to the [RFC7432] mobility procedures.

   An entry may associate a configured static IP to a list of potential
   MACs, i.e., IP1->(MAC1,MAC2..MACN).  Until a frame (including a local
   ARP/NA message) is received from the CE, the PE will not advertise
   any IP1->MAC in EVPN.  Upon receiving traffic from the CE, the PE
   will check that the source MAC, e.g., MAC1, is included in the list
   of allowed MACs.  Only in that case, the PE will activate the
   IP1->MAC1 and advertise only that IP1 and MAC1 in an EVPN MAC/IP
   Advertisement route.

   The PE MUST create EVPN-learned entries from the received valid EVPN
   MAC/IP Advertisement routes containing a MAC and IP address.

   Dynamic entries are learned in different ways depending on whether
   the entry contains an IPv4 or IPv6 address:

   Proxy ARP dynamic entries:
      The PE MUST snoop all ARP packets (that is, all frames with
      Ethertype 0x0806) received from the CEs attached to the BD in
      order to learn dynamic entries.  ARP packets received from remote
      EVPN PEs attached to the same BD are not snooped.  The Learning
      function will add the sender MAC and sender IP of the snooped ARP
      packet to the Proxy ARP table.  Note that a MAC or an IP address
      with value 0 SHOULD NOT be learned.

   Proxy ND dynamic entries:
      The PE MUST snoop the NA messages (Ethertype 0x86dd, ICMPv6 type
      136) received from the CEs attached to the BD and learn dynamic
      entries from the Target Address and TLLA information.  NA messages
      received from remote EVPN PEs are not snooped.  A PE implementing
      Proxy ND as in this document MUST NOT create dynamic IP->MAC
      entries from NS messages because they don't contain the R Flag
      required by the Proxy ND reply function.  See Section 3.2.1 for
      more information about the R Flag.

      This document specifies an "anycast" capability that can be
      configured for the Proxy ND function of the PE and affects how
      dynamic Proxy ND entries are learned based on the O Flag of the
      snooped NA messages.  If the O Flag is zero in the received NA
      message, the IP->MAC SHOULD only be learned in case the IPv6
      "anycast" capability is enabled in the BD.  Irrespective, an NA
      message with O Flag = 0 will be normally forwarded by the PE based
      on a MAC DA lookup.

   The following procedure associated to the Learning sub-function is
   RECOMMENDED:

   *  When a new Proxy ARP/ND EVPN or static active entry is learned (or
      provisioned), the PE SHOULD send a GARP or unsolicited-NA message
      to all the connected access CEs.  The PE SHOULD send a GARP or
      unsolicited-NA message for dynamic entries only if the ARP/NA
      message that previously created the entry on the PE was NOT
      flooded to all the local connected CEs before.  This GARP/
      unsolicited-NA message makes sure the CE ARP/ND caches are updated
      even if the ARP/NS/NA messages from CEs connected to remote PEs
      are not flooded in the EVPN network.

   Note that if a static entry is provisioned with the same IP as an
   existing EVPN-learned or dynamic entry, the static entry takes
   precedence.

   In case of a PE reboot, the static and EVPN entries will be re-added
   as soon as the PE is back online and receives all the EVPN routes for
   the BD.  However, the dynamic entries will be gone.  Due to that
   reason, new NS and ARP Requests will be flooded by the PE to remote
   PEs and dynamic entries gradually relearned again.

3.2.1.  Proxy ND and the NA Flags



   [RFC4861] describes the use of the R Flag in IPv6 address resolution:

   *  Nodes capable of routing IPv6 packets must reply to NS messages
      with NA messages where the R Flag is set (R Flag = 1).

   *  Hosts that are not able to route IPv6 packets must indicate that
      inability by replying with NA messages that contain R Flag = 0.

   The use of the R Flag in NA messages has an impact on how hosts
   select their default gateways when sending packets off-link, as per
   [RFC4861]:

   *  Hosts build a Default Router List based on the received RAs and
      NAs with R Flag = 1.  Each cache entry has an IsRouter flag, which
      must be set for received RAs and is set based on the R Flag in the
      received NAs.  A host can choose one or more Default Routers when
      sending packets off-link.

   *  In those cases where the IsRouter flag changes from TRUE to FALSE
      as a result of an NA update, the node must remove that router from
      the Default Router List and update the Destination Cache entries
      for all destinations using that neighbor as a router, as specified
      in Section 7.3.3 of [RFC4861].  This is needed to detect when a
      node that is used as a router stops forwarding packets due to
      being configured as a host.

   The R and O Flags for a Proxy ARP/ND entry will be learned in the
   following ways:

   *  The R Flag information SHOULD be added to the static entries by
      the management interface.  The O Flag information MAY also be
      added by the management interface.  If the R and O Flags are not
      configured, the default value is 1.

   *  Dynamic entries SHOULD learn the R Flag and MAY learn the O Flag
      from the snooped NA messages used to learn the IP->MAC itself.

   *  EVPN-learned entries SHOULD learn the R Flag and MAY learn the O
      Flag from the ARP/ND Extended Community [RFC9047] received from
      EVPN along with the RT2 used to learn the IP->MAC itself.  If no
      ARP/ND Extended Community is received, the PE will add a
      configured R Flag / O Flag to the entry.  These configured R and O
      Flags MAY be an administrative choice with a default value of 1.
      The configuration of this administrative choice provides a
      backwards-compatible option with EVPN PEs that follow [RFC7432]
      but do not support this specification.

   Note that, typically, IP->MAC entries with O = 0 will not be learned;
   therefore, the Proxy ND function will reply to NS messages with NA
   messages that contain O = 1.  However, this document allows the
   configuration of the "anycast" capability in the BD where the Proxy
   ND function is enabled.  If "anycast" is enabled in the BD and an NA
   message with O = 0 is received, the associated IP->MAC entry will be
   learned with O = 0.  If this "anycast" capability is enabled in the
   BD, duplicate IP detection must be disabled so that the PE is able to
   learn the same IP mapped to different MACs in the same Proxy ND
   table.  If the "anycast" capability is disabled, NA messages with O
   Flag = 0 will not create a Proxy ND entry (although they will be
   forwarded normally); hence, no EVPN advertisement with ARP/ND
   Extended Community will be generated.

3.3.  Reply Sub-function



   This sub-function will reply to address resolution requests/
   solicitations upon successful lookup in the Proxy ARP/ND table for a
   given IP address.  The following considerations should be taken into
   account, assuming that the ARP Request / NS lookup hits a Proxy ARP/
   ND entry IP1->MAC1:

   a.  When replying to ARP Requests or NS messages:

       *  The PE SHOULD use the Proxy ARP/ND entry MAC address MAC1 as
          MAC SA.  This is RECOMMENDED so that the resolved MAC can be
          learned in the MAC forwarding database of potential Layer 2
          switches sitting between the PE and the CE requesting the
          address resolution.

       *  For an ARP reply, the PE MUST use the Proxy ARP entry IP1 and
          MAC1 addresses in the sender Protocol Address and Hardware
          Address fields, respectively.

       *  For an NA message in response to an address resolution NS or
          DAD NS, the PE MUST use IP1 as the IP SA and Target Address.
          M1 MUST be used as the Target Link Local Address (TLLA).

   b.  A PE SHOULD NOT reply to a request/solicitation received on the
       same attachment circuit over which the IP->MAC is learned.  In
       this case, the requester and the requested IP are assumed to be
       connected to the same Layer 2 CE/access network linked to the
       PE's attachment circuit; therefore, the requested IP owner will
       receive the request directly.

   c.  A PE SHOULD reply to broadcast/multicast address resolution
       messages, i.e., ARP Requests, ARP probes, NS messages, as well as
       DAD NS messages.  An ARP probe is an ARP Request constructed with
       an all-zero sender IP address that may be used by hosts for IPv4
       Address Conflict Detection as specified in [RFC5227].  A PE
       SHOULD NOT reply to unicast address resolution requests (for
       instance, NUD NS messages).

   d.  When replying to an NS, a PE SHOULD set the Flags in the NA
       messages as follows:

       *  The R bit is set as it was learned for the IP->MAC entry in
          the NA messages that created the entry (see Section 3.2.1).

       *  The S Flag will be set/unset as per [RFC4861].

       *  The O Flag will be set in all the NA messages issued by the PE
          except in the case in which the BD is configured with the
          "anycast" capability and the entry was previously learned with
          O = 0.  If "anycast" is enabled and there is more than one MAC
          for a given IP in the Proxy ND table, the PE will reply to NS
          messages with as many NA responses as "anycast" entries there
          are in the Proxy ND table.

   e.  For Proxy ARP, a PE MUST only reply to ARP Requests with the
       format specified in [RFC0826].

   f.  For Proxy ND, a PE MUST reply to NS messages with known options
       with the format and options specified in [RFC4861] and MAY reply,
       discard, forward, or unicast-forward NS messages containing other
       options.  An administrative choice to control the behavior for
       received NS messages with unknown options ("reply", "discard",
       "unicast-forward", or "forward") MAY be supported.

       *  The "reply" option implies that the PE ignores the unknown
          options and replies with NA messages, assuming a successful
          lookup on the Proxy ND table.  An unsuccessful lookup will
          result in a "forward" behavior (i.e., flood the NS message
          based on the MAC DA).

       *  If "discard" is available, the operator should assess if
          flooding NS unknown options may be a security risk for the
          EVPN BD (and if so, enable "discard") or, on the contrary, if
          not forwarding/flooding NS unknown options may disrupt
          connectivity.  This option discards NS messages with unknown
          options irrespective of the result of the lookup on the Proxy
          ND table.

       *  The "unicast-forward" option is described in Section 3.4.

       *  The "forward" option implies flooding the NS message based on
          the MAC DA.  This option forwards NS messages with unknown
          options irrespective of the result of the lookup on the Proxy
          ND table.  The "forward" option is RECOMMENDED by this
          document.

3.4.  Unicast-Forward Sub-function



   As discussed in Section 3.3, in some cases, the operator may want to
   "unicast-forward" certain ARP Requests and NS messages as opposed to
   reply to them.  The implementation of a "unicast-forward" function is
   RECOMMENDED.  This option can be enabled with one of the following
   parameters:

   a.  unicast-forward always

   b.  unicast-forward unknown-options

   If "unicast-forward always" is enabled, the PE will perform a Proxy
   ARP/ND table lookup and, in case of a hit, the PE will forward the
   packet to the owner of the MAC found in the Proxy ARP/ND table.  This
   is irrespective of the options carried in the ARP/ND packet.  This
   option provides total transparency in the BD and yet reduces the
   amount of flooding significantly.

   If "unicast-forward unknown-options" is enabled, upon a successful
   Proxy ARP/ND lookup, the PE will perform a "unicast-forward" action
   only if the ARP Requests or NS messages carry unknown options, as
   explained in Section 3.3.  The "unicast-forward unknown-options"
   configuration allows the support of new applications using ARP/ND in
   the BD while still reducing the flooding.

   Irrespective of the enabled option, if there is no successful Proxy
   ARP/ND lookup, the unknown ARP Request / NS message will be flooded
   in the context of the BD, as per Section 3.6.

3.5.  Maintenance Sub-function



   The Proxy ARP/ND tables SHOULD follow a number of maintenance
   procedures so that the dynamic IP->MAC entries are kept if the owner
   is active and flushed (and the associated RT2 withdrawn) or if the
   owner is no longer in the network.  The following procedures are
   RECOMMENDED:

   Age-time:
      A dynamic Proxy ARP/ND entry MUST be flushed out of the table if
      the IP->MAC has not been refreshed within a given age-time.  The
      entry is refreshed if an ARP or NA message is received for the
      same IP->MAC entry.  The age-time is an administrative option, and
      its value should be carefully chosen depending on the specific use
      case; in IXP networks (where the CE routers are fairly static),
      the age-time may normally be longer than in DC networks (where
      mobility is required).

   Send-refresh option:
      The PE MAY send periodic refresh messages (ARP/ND "probes") to the
      owners of the dynamic Proxy ARP/ND entries, so that the entries
      can be refreshed before they age out.  The owner of the IP->MAC
      entry would reply to the ARP/ND probe and the corresponding entry
      age-time reset.  The periodic send-refresh timer is an
      administrative option and is RECOMMENDED to be a third of the age-
      time or a half of the age-time in scaled networks.

      An ARP refresh issued by the PE will be an ARP Request message
      with the sender's IP = 0 sent from the PE's MAC SA.  If the PE has
      an IP address in the subnet, for instance, on an Integrated
      Routing and Bridging (IRB) interface, then it MAY use it as a
      source for the ARP Request (instead of sender's IP = 0).  An ND
      refresh will be an NS message issued from the PE's MAC SA and a
      Link Local Address associated to the PE's MAC.

      The refresh request messages SHOULD be sent only for dynamic
      entries and not for static or EVPN-learned entries.  Even though
      the refresh request messages are broadcast or multicast, the PE
      SHOULD only send the message to the attachment circuit associated
      to the MAC in the IP->MAC entry.

   The age-time and send-refresh options are used in EVPN networks to
   avoid unnecessary EVPN RT2 withdrawals; if refresh messages are sent
   before the corresponding BD Bridge-Table and Proxy ARP/ND age-time
   for a given entry expires, inactive but existing hosts will reply,
   refreshing the entry and therefore avoiding unnecessary EVPN MAC/IP
   Advertisement withdrawals in EVPN.  Both entries (MAC in the BD and
   IP->MAC in the Proxy ARP/ND) are reset when the owner replies to the
   ARP/ND probe.  If there is no response to the ARP/ND probe, the MAC
   and IP->MAC entries will be legitimately flushed and the RT2s
   withdrawn.

3.6.  Flood (to Remote PEs) Handling



   The Proxy ARP/ND function implicitly helps reduce the flooding of ARP
   Requests and NS messages to remote PEs in an EVPN network.  However,
   in certain use cases, the flooding of ARP/NS/NA messages (and even
   the unknown unicast flooding) to remote PEs can be suppressed
   completely in an EVPN network.

   For instance, in an IXP network, since all the participant CEs are
   well known and will not move to a different PE, the IP->MAC entries
   for the local CEs may be all provisioned on the PEs by a management
   system.  Assuming the entries for the CEs are all provisioned on the
   local PE, a given Proxy ARP/ND table will only contain static and
   EVPN-learned entries.  In this case, the operator may choose to
   suppress the flooding of ARP/NS/NA from the local PE to the remote
   PEs completely.

   The flooding may also be suppressed completely in IXP networks with
   dynamic Proxy ARP/ND entries assuming that all the CEs are directly
   connected to the PEs and that they all advertise their presence with
   a GARP/unsolicited-NA when they connect to the network.  If any of
   those two assumptions are not true and any of the PEs may not learn
   all the local Proxy ARP/ND entries, flooding of the ARP/NS/NA
   messages from the local PE to the remote PEs SHOULD NOT be
   suppressed, or the address resolution process for some CEs will not
   be completed.

   In networks where fast mobility is expected (DC use case), it is NOT
   RECOMMENDED
to suppress the flooding of unknown ARP Requests / NS
   messages or GARPs/unsolicited-NAs.  Unknown ARP Requests / NS
   messages refer to those ARP Requests / NS messages for which the
   Proxy ARP/ND lookups for the requested IPs do not succeed.

   In order to give the operator the choice to suppress/allow the
   flooding to remote PEs, a PE MAY support administrative options to
   individually suppress/allow the flooding of:

   *  Unknown ARP Requests and NS messages.

   *  GARP and unsolicited-NA messages.

   The operator will use these options based on the expected behavior on
   the CEs.

3.7.  Duplicate IP Detection



   The Proxy ARP/ND function MUST support duplicate IP detection as per
   this section so that ARP/ND-spoofing attacks or duplicate IPs due to
   human errors can be detected.  For IPv6 addresses, CEs will continue
   to carry out the DAD procedures as per [RFC4862].  The solution
   described in this section is an additional security mechanism carried
   out by the PEs that guarantees IPv6 address moves between PEs are
   legitimate and not the result of an attack.  [RFC6957] describes a
   solution for the IPv6 Duplicate Address Detection Proxy; however, it
   is defined for point-to-multipoint topologies with a split-horizon
   forwarding, where the "CEs" have no direct communication within the
   same L2 link; therefore, it is not suitable for EVPN Broadcast
   Domains.  In addition, the solution described in this section
   includes the use of the AS-MAC for additional security.

   ARP/ND spoofing is a technique whereby an attacker sends "fake" ARP/
   ND messages onto a Broadcast Domain.  Generally, the aim is to
   associate the attacker's MAC address with the IP address of another
   host causing any traffic meant for that IP address to be sent to the
   attacker instead.

   The distributed nature of EVPN and Proxy ARP/ND allows the easy
   detection of duplicated IPs in the network in a similar way to the
   MAC duplication detection function supported by [RFC7432] for MAC
   addresses.

   Duplicate IP detection monitors "IP-moves" in the Proxy ARP/ND table
   in the following way:

   a.  When an existing active IP1->MAC1 entry is modified, a PE starts
       an M-second timer (default value of M = 180), and if it detects N
       IP moves before the timer expires (default value of N = g5), it
       concludes that a duplicate IP situation has occurred.  An IP move
       is considered when, for instance, IP1->MAC1 is replaced by
       IP1->MAC2 in the Proxy ARP/ND table.  Static IP->MAC entries,
       i.e., locally provisioned or EVPN-learned entries with I = 1 in
       the ARP/ND Extended Community, are not subject to this procedure.
       Static entries MUST NOT be overridden by dynamic Proxy ARP/ND
       entries.

   b.  In order to detect the duplicate IP faster, the PE SHOULD send a
       Confirm message to the former owner of the IP.  A Confirm message
       is a unicast ARP Request / NS message sent by the PE to the MAC
       addresses that previously owned the IP, when the MAC changes in
       the Proxy ARP/ND table.  The Confirm message uses a sender's IP
       0.0.0.0 in case of ARP (if the PE has an IP address in the
       subnet, then it MAY use it) and an IPv6 Link Local Address in
       case of NS.  If the PE does not receive an answer within a given
       time, the new entry will be confirmed and activated.  The default
       RECOMMENDED time to receive the confirmation is 30 seconds.  In
       case of spoofing, for instance, if IP1->MAC1 moves to IP1->MAC2,
       the PE may send a unicast ARP Request / NS message for IP1 with
       MAC DA = MAC1 and MAC SA = PE's MAC.  This will force the
       legitimate owner to respond if the move to MAC2 was spoofed and
       make the PE issue another Confirm message, this time to MAC DA =
       MAC2.  If both, the legitimate owner and spoofer keep replying to
       the Confirm message.  The PE would then detect the duplicate IP
       within the M-second timer, and a response would be triggered as
       follows:

       *  If the IP1->MAC1 pair was previously owned by the spoofer and
          the new IP1->MAC2 was from a valid CE, then the issued Confirm
          message would trigger a response from the spoofer.

       *  If it were the other way around, that is, IP1->MAC1 was
          previously owned by a valid CE, the Confirm message would
          trigger a response from the CE.

          Either way, if this process continues, then duplicate
          detection will kick in.

   c.  Upon detecting a duplicate IP situation:

       1.  The entry in duplicate detected state cannot be updated with
           new dynamic or EVPN-learned entries for the same IP.  The
           operator MAY override the entry, though, with a static
           IP->MAC.

       2.  The PE SHOULD alert the operator and stop responding to ARP/
           NS for the duplicate IP until a corrective action is taken.

       3.  Optionally, the PE MAY associate an "anti-spoofing-mac" (AS-
           MAC) to the duplicate IP in the Proxy ARP/ND table.  The PE
           will send a GARP/unsolicited-NA message with IP1->AS-MAC to
           the local CEs as well as an RT2 (with IP1->AS-MAC) to the
           remote PEs.  This will update the ARP/ND caches on all the
           CEs in the BD; hence, all the CEs in the BD will use the AS-
           MAC as MAC DA when sending traffic to IP1.  This procedure
           prevents the spoofer from attracting any traffic for IP1.
           Since the AS-MAC is a managed MAC address known by all the
           PEs in the BD, all the PEs MAY apply filters to drop and/or
           log any frame with MAC DA = AS-MAC.  The advertisement of the
           AS-MAC as a "drop-MAC" (by using an indication in the RT2)
           that can be used directly in the BD to drop frames is for
           further study.

   d.  The duplicate IP situation will be cleared when a corrective
       action is taken by the operator or, alternatively, after a HOLD-
       DOWN timer (default value of 540 seconds).

   The values of M, N, and HOLD-DOWN timer SHOULD be a configurable
   administrative option to allow for the required flexibility in
   different scenarios.

   For Proxy ND, the duplicate IP detection described in this section
   SHOULD only monitor IP moves for IP->MACs learned from NA messages
   with O Flag = 1.  NA messages with O Flag = 0 would not override the
   ND cache entries for an existing IP; therefore, the procedure in this
   section would not detect duplicate IPs.  This duplicate IP detection
   for IPv6 SHOULD be disabled when the IPv6 "anycast" capability is
   activated in a given BD.

4.  Solution Benefits

   The solution described in this document provides the following
   benefits:

   a.  May completely suppress the flooding of the ARP/ND messages in
       the EVPN network, assuming that all the CE IP->MAC addresses
       local to the PEs are known or provisioned on the PEs from a
       management system.  Note that in this case, the unknown unicast
       flooded traffic can also be suppressed, since all the expected
       unicast traffic will be destined to known MAC addresses in the PE
       BDs.

   b.  Significantly reduces the flooding of the ARP/ND messages in the
       EVPN network, assuming that some or all the CE IP->MAC addresses
       are learned on the data plane by snooping ARP/ND messages issued
       by the CEs.

   c.  Provides a way to refresh periodically the CE IP->MAC entries
       learned through the data plane so that the IP->MAC entries are
       not withdrawn by EVPN when they age out unless the CE is not
       active anymore.  This option helps reducing the EVPN control
       plane overhead in a network with active CEs that do not send
       packets frequently.

   d.  Provides a mechanism to detect duplicate IP addresses and avoid
       ARP/ND-spoof attacks or the effects of duplicate addresses due to
       human errors.

5.  Deployment Scenarios

   Four deployment scenarios with different levels of ARP/ND control are
   available to operators using this solution depending on their
   requirements to manage ARP/ND: all dynamic learning, all dynamic
   learning with Proxy ARP/ND, hybrid dynamic learning and static
   provisioning with Proxy ARP/ND, and all static provisioning with
   Proxy ARP/ND.

5.1.  All Dynamic Learning



   In this scenario for minimum security and mitigation, EVPN is
   deployed in the BD with the Proxy ARP/ND function shutdown.  PEs do
   not intercept ARP/ND requests and flood all requests issued by the
   CEs as a conventional Layer 2 network among those CEs would suffice.
   While no ARP/ND mitigation is used in this scenario, the operator can
   still take advantage of EVPN features such as control plane learning
   and all-active multihoming in the peering network.

   Although this option does not require any of the procedures described
   in this document, it is added as a baseline/default option for
   completeness.  This option is equivalent to VPLS as far as ARP/ND is
   concerned.  The options described in Sections 5.2, 5.3, and 5.4 are
   only possible in EVPN networks in combination with their Proxy ARP/ND
   capabilities.

5.2.  Dynamic Learning with Proxy ARP/ND



   This scenario minimizes flooding while enabling dynamic learning of
   IP->MAC entries.  The Proxy ARP/ND function is enabled in the BDs of
   the EVPN PEs so that the PEs snoop ARP/ND messages issued by the CEs
   and respond to CE ARP Requests / NS messages.

   PEs will flood requests if the entry is not in their Proxy table.
   Any unknown source IP->MAC entries will be learned and advertised in
   EVPN, and traffic to unknown entries is discarded at the ingress PE.

   This scenario makes use of the Learning, Reply, and Maintenance sub-
   functions, with an optional use of the Unicast-forward and duplicate
   IP detection sub-functions.  The Flood handling sub-function uses
   default flooding for unknown ARP Requests / NS messages.

5.3.  Hybrid Dynamic Learning and Static Provisioning with Proxy ARP/ND



   Some IXPs and other operators want to protect particular hosts on the
   BD while allowing dynamic learning of CE addresses.  For example, an
   operator may want to configure static IP->MAC entries for management
   and infrastructure hosts that provide critical services.  In this
   scenario, static entries are provisioned from the management plane
   for protected IP->MAC addresses, and dynamic learning with Proxy ARP/
   ND is enabled as described in Section 5.2 on the BD.

   This scenario makes use of the same sub-functions as in Section 5.2
   but with static entries added by the Learning sub-function.

5.4.  All Static Provisioning with Proxy ARP/ND



   For a solution that maximizes security and eliminates flooding and
   unknown unicast in the peering network, all IP->MAC entries are
   provisioned from the management plane.  The Proxy ARP/ND function is
   enabled in the BDs of the EVPN PEs so that the PEs intercept and
   respond to CE requests.  Dynamic learning and ARP/ND snooping is
   disabled so that ARP Requests and NS messages to unknown IPs are
   discarded at the ingress PE.  This scenario provides an operator the
   most control over IP->MAC entries and allows an operator to manage
   all entries from a management system.

   In this scenario, the Learning sub-function is limited to static
   entries, the Maintenance sub-function will not require any procedures
   due to the static entries, and the Flood handling sub-function will
   completely suppress unknown ARP Requests / NS messages as well as
   GARP and unsolicited-NA messages.

5.5.  Example of Deployment in Internet Exchange Points



   Nowadays, almost all IXPs install some security rules in order to
   protect the peering network (BD).  These rules are often called port
   security.  Port security summarizes different operational steps that
   limit the access to the IXP-LAN and the customer router and controls
   the kind of traffic that the routers are allowed to exchange (e.g.,
   Ethernet, IPv4, and IPv6).  Due to this, the deployment scenario as
   described in Section 5.4, "All Static Provisioning with Proxy ARP/
   ND", is the predominant scenario for IXPs.

   In addition to the "All Static Provisioning" behavior, in IXP
   networks it is recommended to configure the Reply sub-function to
   "discard" ARP Requests / NS messages with unrecognized options.

   At IXPs, customers usually follow a certain operational life cycle.
   For each step of the operational life cycle, specific operational
   procedures are executed.

   The following describes the operational procedures that are needed to
   guarantee port security throughout the life cycle of a customer with
   focus on EVPN features:

   1.  A new customer is connected the first time to the IXP:

       Before the connection between the customer router and the IXP-LAN
       is activated, the MAC of the router is allowlisted on the IXP's
       switch port.  All other MAC addresses are blocked.  Pre-defined
       IPv4 and IPv6 addresses of the IXP peering network space are
       configured at the customer router.  The IP->MAC static entries
       (IPv4 and IPv6) are configured in the management system of the
       IXP for the customer's port in order to support Proxy ARP/ND.

       In case a customer uses multiple ports aggregated to a single
       logical port (LAG), some vendors randomly select the MAC address
       of the LAG from the different MAC addresses assigned to the
       ports.  In this case, the static entry will be used and
       associated to a list of allowed MACs.

   2.  Replacement of customer router:

       If a customer router is about to be replaced, the new MAC
       address(es) must be installed in the management system in
       addition to the MAC address(es) of the currently connected
       router.  This allows the customer to replace the router without
       any active involvement of the IXP operator.  For this, static
       entries are also used.  After the replacement takes place, the
       MAC address(es) of the replaced router can be removed.

   3.  Decommissioning a customer router:

       If a customer router is decommissioned, the router is
       disconnected from the IXP PE.  Right after that, the MAC
       address(es) of the router and IP->MAC bindings can be removed
       from the management system.

5.6.  Example of Deployment in Data Centers



   DCs normally have different requirements than IXPs in terms of Proxy
   ARP/ND.  Some differences are listed below:

   a.  The required mobility in virtualized DCs makes the "Dynamic
       Learning" or "Hybrid Dynamic and Static Provisioning" models more
       appropriate than the "All Static Provisioning" model.

   b.  IPv6 "anycast" may be required in DCs, while it is typically not
       a requirement in IXP networks.  Therefore, if the DC needs IPv6
       anycast addresses, the "anycast" capability will be explicitly
       enabled in the Proxy ND function and hence the Proxy ND sub-
       functions modified accordingly.  For instance, if IPv6 "anycast"
       is enabled in the Proxy ND function, the duplicate IP detection
       procedure in Section 3.7 must be disabled.

   c.  DCs may require special options on ARP/ND as opposed to the
       address resolution function, which is the only one typically
       required in IXPs.  Based on that, the Reply sub-function may be
       modified to forward or discard unknown options.

6.  Security Considerations

   The security considerations of [RFC7432] and [RFC9047] apply to this
   document too.  Note that EVPN does not inherently provide
   cryptographic protection (including confidentiality protection).

   The procedures in this document reduce the amount of ARP/ND message
   flooding, which in itself provides a protection to "slow path"
   software processors of routers and Tenant Systems in large BDs.  The
   ARP/ND requests that are replied to by the Proxy ARP/ND function
   (hence not flooded) are normally targeted to existing hosts in the
   BD.  ARP/ND requests targeted to absent hosts are still normally
   flooded; however, the suppression of unknown ARP Requests and NS
   messages described in Section 3.6 can provide an additional level of
   security against ARP Requests / NS messages issued to non-existing
   hosts.

   While the unicast-forward and/or flood suppression sub-functions
   provide an added security mechanism for the BD, they can also
   increase the risk of blocking the service for a CE if the EVPN PEs
   cannot provide the ARP/ND resolution that the CE needs.

   The solution also provides protection against Denial-of-Service (DoS)
   attacks that use ARP/ND spoofing as a first step.  The duplicate IP
   detection and the use of an AS-MAC as explained in Section 3.7
   protects the BD against ARP/ND spoofing.

   The Proxy ARP/ND function specified in this document does not allow
   for the learning of an IP address mapped to multiple MAC addresses in
   the same table unless the "anycast" capability is enabled (and only
   in case of Proxy ND).  When "anycast" is enabled in the Proxy ND
   function, the number of allowed entries for the same IP address MUST
   be limited by the operator to prevent DoS attacks that attempt to
   fill the Proxy ND table with a significant number of entries for the
   same IP.

   This document provides some examples and guidelines that can be used
   by IXPs in their EVPN BDs.  When EVPN and its associated Proxy ARP/ND
   function are used in IXP networks, they provide ARP/ND security and
   mitigation.  IXPs must still employ additional security mechanisms
   that protect the peering network as per the established BCPs such as
   the ones described in [EURO-IX-BCP].  For example, IXPs should
   disable all unneeded control protocols and block unwanted protocols
   from CEs so that only IPv4, ARP, and IPv6 Ethertypes are permitted on
   the peering network.  In addition, port security features and ACLs
   can provide an additional level of security.

   Finally, it is worth noting that the Proxy ARP/ND solution in this
   document will not work if there is a mechanism securing ARP/ND
   exchanges among CEs because the PE is not able to secure the
   "proxied" ND messages.

7.  IANA Considerations



   This document has no IANA actions.

8.  References



8.1.  Normative References



   [RFC0826]  Plummer, D., "An Ethernet Address Resolution Protocol: Or
              Converting Network Protocol Addresses to 48.bit Ethernet
              Address for Transmission on Ethernet Hardware", STD 37,
              RFC 826, DOI 10.17487/RFC0826, November 1982,
              <https://www.rfc-editor.org/info/rfc826>.

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <https://www.rfc-editor.org/info/rfc2119>.

   [RFC4861]  Narten, T., Nordmark, E., Simpson, W., and H. Soliman,
              "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861,
              DOI 10.17487/RFC4861, September 2007,
              <https://www.rfc-editor.org/info/rfc4861>.

   [RFC5227]  Cheshire, S., "IPv4 Address Conflict Detection", RFC 5227,
              DOI 10.17487/RFC5227, July 2008,
              <https://www.rfc-editor.org/info/rfc5227>.

   [RFC7432]  Sajassi, A., Ed., Aggarwal, R., Bitar, N., Isaac, A.,
              Uttaro, J., Drake, J., and W. Henderickx, "BGP MPLS-Based
              Ethernet VPN", RFC 7432, DOI 10.17487/RFC7432, February
              2015, <https://www.rfc-editor.org/info/rfc7432>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <https://www.rfc-editor.org/info/rfc8174>.

   [RFC9047]  Rabadan, J., Ed., Sathappan, S., Nagaraj, K., and W. Lin,
              "Propagation of ARP/ND Flags in an Ethernet Virtual
              Private Network (EVPN)", RFC 9047, DOI 10.17487/RFC9047,
              June 2021, <https://www.rfc-editor.org/info/rfc9047>.

8.2.  Informative References



   [EURO-IX-BCP]
              Euro-IX, "European Internet Exchange Association",
              <https://www.euro-ix.net/en/forixps/set-ixp/ixp-bcops>.

   [RFC4862]  Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless
              Address Autoconfiguration", RFC 4862,
              DOI 10.17487/RFC4862, September 2007,
              <https://www.rfc-editor.org/info/rfc4862>.

   [RFC6820]  Narten, T., Karir, M., and I. Foo, "Address Resolution
              Problems in Large Data Center Networks", RFC 6820,
              DOI 10.17487/RFC6820, January 2013,
              <https://www.rfc-editor.org/info/rfc6820>.

   [RFC6957]  Costa, F., Combes, J-M., Ed., Pougnard, X., and H. Li,
              "Duplicate Address Detection Proxy", RFC 6957,
              DOI 10.17487/RFC6957, June 2013,
              <https://www.rfc-editor.org/info/rfc6957>.

Acknowledgments

   The authors want to thank Ranganathan Boovaraghavan, Sriram
   Venkateswaran, Manish Krishnan, Seshagiri Venugopal, Tony Przygienda,
   Robert Raszuk, and Iftekhar Hussain for their review and
   contributions.  Thank you to Oliver Knapp as well for his detailed
   review.

Contributors

   In addition to the authors listed on the front page, the following
   coauthors have also contributed to this document:

   Wim Henderickx
   Nokia


   Daniel Melzer
   DE-CIX Management GmbH


   Erik Nordmark
   Zededa


Authors' Addresses



   Jorge Rabadan (editor)
   Nokia
   777 Middlefield Road
   Mountain View, CA 94043
   United States of America

   Email: jorge.rabadan@nokia.com


   Senthil Sathappan
   Nokia
   701 E. Middlefield Road
   Mountain View, CA 94043
   United States of America

   Email: senthil.sathappan@nokia.com


   Kiran Nagaraj
   Nokia
   701 E. Middlefield Road
   Mountain View, CA 94043
   United States of America

   Email: kiran.nagaraj@nokia.com


   Greg Hankins
   Nokia

   Email: greg.hankins@nokia.com


   Thomas King
   DE-CIX Management GmbH

   Email: thomas.king@de-cix.net