Network working group                                             L. Xia
Internet Draft                                                   L. Yong
Category: Standard Track                                            Weiguo Hao
                                                          Anoop Ghanwani
                                                            Ram Krishnan
Expires: April 2015                                     October 27, 2014

                          Layer 2 Gateway (L2GW)


   A Layer 2 Gateway (L2GW) is used for interconnecting a Layer 2
   overlay network [NVO3FRWK] and a Layer 2 bridged network [IEEE802.1Q]
   to form a single Layer 2 virtual network.  This draft describes data
   plane interconnection and control plane interworking at the L2GW.

Status of this Memo

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

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups. Note that
   other groups may also distribute working documents as Internet-

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

   The list of current Internet-Drafts can be accessed at

   The list of Internet-Draft Shadow Directories can be accessed at

   This Internet-Draft will expire on April 27, 2015.

Xia, et al.                                                  [Page 1]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

Copyright Notice

   Copyright (c) 2013 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
   ( 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.

Table of Contents

   1. Introduction ................................................ 3
      1.1. Conventions used in this document ...................... 3
      1.2. Terminology ............................................ 3
   2. L2GW Reference Model......................................... 4
   3. General L2GW Operation Procedures ........................... 5
      3.1. MAC Learning ........................................... 5
      3.2. ARP Handling ........................................... 6
      3.3. Dual L2GWs ............................................. 6
   4. L2CP Review and Applicability to L2 Overlay Network ......... 8
      4.1. STP/RSTP/MSTP ......................................... 10
      4.2. PAUSE ................................................. 11
      4.3. LACP/LAMP ............................................. 11
      4.4. Link OAM .............................................. 12
      4.5. Port Authentication ................................... 12
      4.6. E-LMI ................................................. 13
      4.7. LLDP .................................................. 13
      4.8. PTP Peer Delay ........................................ 13
      4.9. ESMC .................................................. 13
      4.10. GARP/MRP Block........................................ 14
   5. L2CP Processing in L2GWs ................................... 14
      5.1. L2CP Frames Filtered (Peered or Discarded) in L2GW .... 14
      5.2. L2CP Frames Passed through L2GW ....................... 15
   6. Other Interworking Cases ................................... 15
   7. Security Considerations .................................... 16
   8. IANA Considerations ........................................ 16
   9. References ................................................. 16
      9.1. Normative References .................................. 16
      9.2. Informative References ................................ 16

Xia, et al.                                                  [Page 2]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

1. Introduction

   Cloud computing and network virtualization are evolving in the
   direction of using network virtualization overlays over Layer 3
   (NVO3).  Some of the goals of NVO3 are -- fast and easy creation of
   tenant networks, support tenant system mobility, and improved
   manageability of all virtualized resources in the data center (DC).

   Layer 2 (L2) overlay network in NVO3 means tenant systems are
   interconnected at L2, while the NVEs are interconnected using Layer
   3 (L3). As a result, it forms a full mesh topology of overlay
   network, i.e. only one L2 hop between any pair of NVEs. On the other
   hand, L2 bridged network is used to refer to the L2 network as
   specified in IEEE 802.1Q [IEEE 802.1Q] in this draft.

   In the first use case, involving DC network migration from physical
   tenant systems to virtual tenant systems, it is expected that the L2
   overlay network may be used along with an existing L2 bridged
   network in a DC, and communication between them would be required.
   In the last use case, a L2 bridged network would be used to connect
   physical (non-virtualized) systems. These devices need to
   communicate to virtualized networks for information exchange. Some
   CPU-intensive applications such as big data analytics typically use
   physical servers rather than making of use of server virtualization.

   To interconnect two networks that are implemented with different
   technologies (NVO3 and a bridged network), gateway functions are
   needed on the device(s)/system(s) that interconnect them.  This
   device is referred to as a Layer 2 Gateway (L2GW) in this draft. The
   device can be thought of as implementing an NVE that connects the
   tenant systems in the L2 bridged network to tenant systems in the
   NVO3 network.

    1.1. Conventions used in this document

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in RFC-2119 [RFC2119].

    1.2. Terminology

   This document uses the terms defined in NVO3 framework [NVO3FRWK]
   and architecture [NVO3ARCH] documents.

Xia, et al.                                                  [Page 3]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

2. L2GW Reference Model

   The following figure shows a reference model where an L2GW provides
   an interconnection between an L2 overlay network and an L2 bridged
   network.  It shows the case where two different technologies are
   used to implement a single L2 network.

                .........                   .........
           +---+         ...           ....          . +------+
       TSs-+NVE|            +---------+              +-+Server|
           +---+ L2 Overlay |         |  L2 Bridge   . +------+
             .    Network   |  L2GW   |   Network    .
             .              |         |              . +------+
         ..+---+            +---------+              +-+Server|
       TSs-+NVE|         ...           ....        ... +------+
           +---+.........                  ........

                      Figure 1: L2GW Reference Model

   The L2GW can reside at the edge of the network providing direct
   connection to tenant systems, or reside at aggregation or core where
   the tenant systems attach to L2 switches. To connect with an L2
   overlay network, an L2GW device physically connects to the underlay
   network on which the L2 overlay network is implemented and it
   functions as an NVE, providing termination for the L2 overlay
   network .

   To provide node failure resilience, the reference model can further
   be shown as in Figure 2, where two L2GWs interconnect the two

                .........                   .........
           +---+         ...           ....          . +------+
       TSs-+NVE|            +---------+              +-+Server|
           +---+ L2 Overlay |  L2GW   |  L2 Bridge   . +------+
             .    Network   +---------+   Network    .
             .                                       . +------+
         ..+---+            +---------+              +-+Server|
       TSs-+NVE|         ...|  L2GW   |....        ... +------+
           +---+.........   +---------+    ........

                      Figure 2: Redundant L2GW Model

   Note that this draft assumes that L2GW device embeds an L2 NVE as
   well as IEEE802.1Q bridge functions.

Xia, et al.                                                  [Page 4]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

3. General L2GW Operation Procedures

    3.1. MAC Learning

   The MAC addresses for an L2 virtual network created by
   interconnecting the two networks (the L2 overlay network and the L2
   bridged network) needs to be distributed and/or learned at all NVEs
   that participate in that L2 virtual network. If NVE-NVA architecture
   is used, when an L2GW learns the MAC addresses from the bridged
   network, the L2GW should notify NVA of the MAC addresses. The NVA
   maintains the mapping of these MAC addresses from the L2GW, and
   informs the other NVEs of the mappings.

   Similarly, if the NVA maintains the mappings between a tenant
   system's MAC address and NVE for an L2 virtual network, the NVA
   would be expected to inform those mappings of MAC addresses to NVEs
   to the L2GWs because the L2GWs also implement the functions of an
   NVE.  The L2GW maintains the mapping of VNID from the L2 overlay
   network and VLAN ID in the bridged network. These mappings may be
   manually configured at the L2GW or may be configured via the NVA.

   The L2GW maintains a forwarding table per virtual network which has
   all the MAC addresses learned from the bridged network as well as
   all of the MAC addresses it received from the NVA for that virtual

   Upon receiving a packet from the overlay network, the L2GW
   decapsulates the packet, performs the table lookup, and may insert a
   VLAN ID (if the decapsulated frame doesn't already have one) or
   modify the VLAN ID (if one is already present) prior to forwarding
   it to the bridged network. If the destination MAC address of the
   decapsulated packet is unknown (i.e. not present in the forwarding
   table), the L2GW may choose to discard the packet or flood it on the
   VLAN depending on the configured policy.

   Upon receiving a frame from the L2 bridge network, the L2GW
   encapsulates the frame prior to forwarding it to the remote NVE. If
   the frame's MAC DA is unknown to L2GW, it will be discarded or
   flooded to all the remote NVEs depending on the configured policy.
   Note that the outer VLAN ID on the packet may be removed before the

   The two networks which are interconnected to form a single L2
   virtual network MUST NOT have any overlapping MAC addresses; i.e.
   the same MAC address cannot appear in the both the L2 overlay
   network as well as the L2 bridged network.

Xia, et al.                                                  [Page 5]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

    3.2. ARP Handling

   To avoid ARP flooding in the L2 overlay network, the L2GW may
   maintain an ARP cache locally and/or rely on NVA to maintain the ARP
   table. For the purpose of maintaining the ARP cache locally, the
   L2GW can snoop ARP requests from the bridged network and send ARP
   replies back.

   If the L2 overlay network supports ARP flooding, the L2GW can simply
   flood ARP requests from one network to another.

    3.3. Dual L2GWs

   Two L2GWs may be used for network interconnection to support a
   network that is resilient to node failures. These two L2GWs may
   further operate in Active/Standby or Active/Active mode. In
   Active/Standby mode, only one of the L2GWs is actively passing
   traffic from one network to the other for a given L2 virtual network.
   In Active/Active mode, both L2GWs pass traffic from one network to
   the other for a given virtual network.

   (TBD: Does this need to be restricted to only two L2GWs?)

   In Active/Standby mode, to protect node failure, some protocol is
   necessary between the L2GWs to facilitate status exchange and
   determine which of them will operate in Active mode. The
   Active/Standby role may be configured or automatically selected
   based on an algorithm or policy. An L2GW should inform NVA about its
   role, i.e., Active or Standby, and the NVA should ensure that the
   active L2GW IP address is used in the mapping of (inner) MAC
   addresses to (outer) IP address.

   In Active/Active mode, NVA/NVEs have two paths to the bridge network
   and vise versa. The NVEs in an overlay can choose one based on the

   The following presents the problems that need to be addressed and
   related solutions for Active/Active connection scenarios:

      1. MAC flip-flop on remote NVEs

   MAC learning on an L2GW can be performed either in data plane or
   control plane. When a local host h1 attaches to multiple L2GWs,
   address learning at the remote NVEs for a given host h1 may
   experience what we refer to as the MAC flip-flop problem where h1
   appears behind the NVE of one L2GW and then subsequently appears

Xia, et al.                                                  [Page 6]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   behind the NVE of the other L2GW, going back and forth in this

   In the data plane learning scenario, an anycast L2GW IP address that
   is shared among L2GWs may be used to avoid MAC flip-flop on remote
   devices (NVEs, L2GWs, etc). When a bridged network attaches to
   multiple L2GWs, any L2GW should use the shared anycast IP address,
   rather than its own IP address, as the ingress NVE IP address when
   it forwards NVO3 data frames into overlay network.  Use of an
   anycast L2GW IP address makes the MAC addresses learnt by the remote
   devices appear to be behind a single source IP address rather than
   multiple different source IP addresses.

   In the control plane learning scenario (i.e. when NVA-NVE is used to
   learn address mappings), if an L2 bridged network is multi-homed to
   multiple L2GWs in Active/Active mode, each edge L2GW should announce
   the MAC addresses of its attached end systems to all other devices
   through NVE-NVA control plane protocol. For MAC addresses that
   originate from multiple L2GWs, remote devices will learn the MAC
   addresses as being associated with multiple ingress IP addresses and
   will generate multiple MAC forwarding entries in ECMP mode. All edge
   L2GWs should disable the data plane MAC learning function in their
   NVEs; they must still continue to learn MAC addresses from traffic
   received from the L2 bridged network. MAC address to NVE IP address
   association should be learned only through the control plane.  The
   control plane must be aware of edge ports that are multi-homed to
   multiple L2GWs.

      2. Duplicated traffic from remote device

   Frame duplication may occur when BUM (broadcast, unknown unicast,
   multicast) traffics are forwarded bidirectionally between an L2
   bridged network and a NVO3 network which have an Active/Active
   connection through multiple edge L2GWs. The Designated Forwarder (DF)
   election mechanism defined in [EVPN] can be used to resolve this
   issue. According to [EVPN], multi-homing functions cover two
   scenarios. For the MHN (Multi-Homed Network) scenario, DF election
   mechanism allows only one L2GW of an edge group to forward BUM
   traffics between NVO3 network and the L2 bridged network by two
   directions for each VN. The basic idea of DF is to elect one L2GW
   per VN from an edge group to be responsible for forwarding the BUM
   traffics.  For the MHD (Multi-Homed Device) scenario, the only
   difference with MHN scenario is at the L2 bridged network side, MC-
   LAG mechanism guarantees BUM traffics coming from L2 bridged network
   only goes to one L2GW. DF mechanism is not needed in this direction.

      3. Loops

Xia, et al.                                                  [Page 7]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   Consider the case where a bridged network originates a frame that is
   sent as a BUM frame to the NVO3 network via an L2GW, say L2GW1, that
   is one of multiple gateways interconnecting the bridged network and
   the NVO3 network. This frame will be encapsulated and then forwarded
   through NVO3 network and reach the other L2GW, say L2GW2, that is
   also connected to the bridged network. In this case, if L2GW2
   decapsulates the NVO3 frame and forwards it into the bridged network
   where the frame originated, the frame loops endlessly. This is why
   it is important to have only single designated forwarder for
   multicast traffic.

      4. Unsynchronized information among member L2GWs

   A local L2GW, say L2GW1 in an edge group, may have learned a VLAN
   and MAC to IP correspondence for a remote end system ES1 when ES1
   sends a packet to local bridge. The returning traffic from local
   bridge may go to any other member L2GW of MC-LAG, for example L2GW2.
   To avoid flooding unicast traffic on L2GW2, MAC address should be
   synchronized among the edge L2GWs in an edge group.

   Additionally, to ensure DF election consistency, dynamic joined VLAN
   through VLAN registration protocol (VRP, [IEEE 802.1ak] amendment to
   the [IEEE 802.1Q]) and dynamic joined multicast group through IGMP
   or MLD protocol should be synchronized among all L2GWs in an edge

4. L2CP Review and Applicability to L2 Overlay Network

   This Section mainly discusses which L2CP (Layer 2 Control Protocol,
   specified in [IEEE 802.1Q]) should be supported by L2 overlay
   network and which should not, Section 5 specifies how L2GW should
   deal with L2CP frames.

   L2CP protocols defined in [IEEE 802.1Q] are listed in Table 1:

   |MAC DA            |Assignment| Protocol |      L2CP Action    |
   |                  |          |  Type    +----------+----------+
   |                  |          |          |VLAN-based|PORT-based|
   |                  |          |          |    L2    |    L2    |
   |                  |          |          | services | services |
   |01-80-C2-00-00-00 |Nearest   |STP/RSTP/M|Filter    |Pass      |
   |                  |Customer  |STP,      |          |          |
   |                  |Bridge    |LACP/LAMP |          |          |

Xia, et al.                                                  [Page 8]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   |01-80-C2-00-00-01 |IEEE MAC  |PAUSE     |Filter    |Filter    |
   |                  |Specific  |          |          |          |
   |                  |Control   |          |          |          |
   |                  |Protocols |          |          |          |
   |01-80-C2-00-00-02 |IEEE 802  |LACP/LAMP,|Filter    |Filter    |
   |                  |Slow      |Link OAM, |          |          |
   |                  |Protocols |ESMC      |          |          |
   |01-80-C2-00-00-03 |Nearest   |Port      |Filter    |Filter    |
   |                  |non-TPRM  |Authentica|          |          |
   |                  |Bridge    |tion,     |          |          |
   |                  |          |LACP/LAMP |          |          |
   |01-80-C2-00-00-04 |IEEE MAC  |          |Filter    |Filter    |
   |                  |Specific  |          |          |          |
   |                  |Control   |          |          |          |
   |                  |Protocols |          |          |          |
   |01-80-C2-00-00-05 |Reserved  |          |Filter    |Filter    |
   |                  |for Future|          |          |          |
   |01-80-C2-00-00-06 |Standardiz|          |          |          |
   |                  |ation     |          |          |          |
   |01-80-C2-00-00-09 |          |          |          |          |
   |                  |          |          |          |          |
   |01-80-C2-00-00-0A |          |          |          |          |
   |01-80-C2-00-00-07 |MEF ELMI  |E-LMI     |Filter    |Filter    |
   |01-80-C2-00-00-08 |Provide   |          |Filter    |Filter    |
   |                  |Bridge    |          |          |          |
   |                  |Group     |          |          |          |
   |01-80-C2-00-00-0B |Reserved  |          |Filter    |Pass      |
   |                  |for Future|          |          |          |
   |01-80-C2-00-00-0C |Standardiz|          |          |          |
   |                  |ation     |          |          |          |
   |01-80-C2-00-00-0D |Provider  |          |Filter    |Pass      |
   |                  |Bridge    |          |          |          |
   |                  |MVRP      |          |          |          |

Xia, et al.                                                  [Page 9]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   |01-80-C2-00-00-0E |Nearest   |LLDP, PTP |Filter    |Filter    |
   |                  |Bridge,   |Peer Delay|          |          |
   |                  |Individual|          |          |          |
   |                  |LAN Scope |          |          |          |
   |01-80-C2-00-00-20 |          |GARP/MRP  |Pass      |Pass      |
   |                  |          |Block     |          |          |
   |      through     |          |          |          |          |
   |                  |          |          |          |          |
   |01-80-C2-00-00-2F |          |          |          |          |

                   Table 1 L2CP protocols specification


      Different L2CP protocols can use the same MAC DA in above block of
      32 addresses, but be differentiated by protocol identifier. MAC DA
      determines the intended recipient device for the frame;

      Filter represent the L2CP action of peer or discard;

      Based on whether L2 interface is VLAN-aware, L2 services can
      divided into two categories: VLAN-based L2 services, PORT-based L2
      services. L2CP action (peer, discard, pass) for these two L2
      services is also different;

      Whether the L2CP frames are peered or discarded is further
      determined by the configuration of L2 interface.

   Further analysis about whether a L2CP protocol is necessary and how
   it is processed in NVO3 supported L2 VN, is provided in the
   following sub sections.

    4.1. STP/RSTP/MSTP

   The Spanning Tree Protocol (STP) is a L2 protocol that ensures a
   loop-free topology for any bridged Ethernet local area network.  The
   basic function of STP is to prevent bridge loops and the broadcast
   storm that results from them. Rapid spanning Tree Protocol (RSTP)
   and Multiple Spanning Tree Protocol (MSTP) are all the enhanced xSTP

   L2 overlay network does not need xSTP protocols to prevent bridge
   loops because it has its own mechanism for it, i.e., NVA, control

Xia, et al.                                                 [Page 10]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   plane mechanisms, full mesh + split horizon, etc. So, the process of
   xSTP frames in L2 VN is:

      Be in line with L2CP protocols' specification of Table 1 from IEEE
      in the L2 sub-networks attached to L2 NVEs;

      xSTP frames are filtered in L2 NVEs and should not go into L2
      overlay network.

    4.2. PAUSE

   [IEEE 802.3-2005] has specified a L2 flow control mechanism through
   using the PAUSE frame. This frame uses L2CP MAC DA of 01-80-C2-00-
   00-01 to be sent to the node at the other end of the link for
   informing it to halt the frame transmission for a specified period
   of time.

   When L2 NVE is co-located in Hypervisor, PAUSE frame is not
   necessary in one device. When they are separated, PAUSE frame is
   only used in layer 2 network between L2 NVE and Hypervisor, there is
   no need to overlay PAUSE frame between L2 NVEs. For the underlay
   network of NVO3 network, L2 PAUSE mechanism is still used between
   two adjacent switches for flow control.

    4.3. LACP/LAMP

   Link Aggregation [IEEE 802.1AXbk-2012] is a mechanism for making
   multiple point-to-point links between a pair of devices appear to be
   a single logical link between those devices. Link Aggregation
   Control Protocol (LACP) and Link Marker Control Protocol (LAMP)
   operate between exactly two peer devices for the purpose of creating,
   verifying, and monitoring the logical link created by aggregating
   individual links.  Specific L2CP frames, known as Link Aggregation
   Control Protocol Data Units (LACPDUs), are exchanged between the
   peer devices on each individual link in the aggregation.  The
   protocol identifier used by LACP is an Ethertype with a value of
   0x8809 (the ''Slow Protocols'' Ethertype) and subtype values 01 (for
   LACP) and 02 (for LAMP). Note that LACP is used to represent LACP
   and LAMP in the following text.

   LACP uses 3 different L2CP MAC DAs to determine the scope of
   propagation of LACPDUs within a bridged LAN, as Table 2 follows:

   |Assignment      | L2CP MAC DA      |Peered or discarded by       |

Xia, et al.                                                 [Page 11]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   |Nearest Customer| 01-80-C2-00-00-00|End Station, Customer Bridge,|
   |Bridge          |                  |Provider Edge Bridge         |
   |IEEE 802 Slow   | 01-80-C2-00-00-02|End Station, Customer Bridge,|
   |Protocols       |                  |Provider Edge Bridge,        |
   |                |                  |Provider Bridge              |
   |Nearest non-TPRM| 01-80-C2-00-00-03|Bridges except for Two Port  |
   |Bridge          |                  |MAC Relay                    |
                Table 2 LACP specification of L2CP MAC DAs

   Base on the summary of Table 2, LACPDUs with the L2CP MAC DA of 01-
   80-C2-00-00-02 are peered or discarded by every node, so this kind
   of LACPDUs will not be overlaid across the L2 overlay network. For
   01-80-C2-00-00-00, it is possible that LACPDUs need to be overlaid
   across Provider Bridge and L2 NVEs of L2 overlay network to reach
   the other end Custom Bridge, L2 overlay network maybe need to
   support to overlay this kind of LACP frame between L2 NVEs. How the
   L2 overlay network support LACP frame of 01-80-C2-00-00-03 is TBD.

    4.4. Link OAM

   Lin OAM defined is defined in [IEEE 802.3ah], as mechanisms for
   monitoring and troubleshooting Ethernet access links. Specifically
   it defines tools for discovery, remote failure indication, remote
   and local loopbacks and status and performance monitoring.

   The Link OAM frames using L2CP MAC DA of 01-80-C2-00-00-02 are
   peered or discarded by every node, so this kind of frame will not be
   overlaid across the L2 overlay network.

    4.5. Port Authentication

   [IEEE 802.1X] is an IEEE Standard for Port-based Network Access
   Control (PNAC). It is part of the IEEE 802.1 group of networking
   protocols. It provides an authentication mechanism to devices
   wishing to attach to a LAN or WLAN.

   Whether or not the L2 overlay network needs to overlay this L2CP
   frames is TBD.

Xia, et al.                                                 [Page 12]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

    4.6. E-LMI

   Ethernet Local Management Interface (E-LMI) [MEF-16] is a protocol
   between the customer edge (CE) device and the provider edge (PE)
   device. It runs only on the PE-CE UNI link and notifies the CE of
   connectivity status and configuration parameters of Ethernet
   services available on the CE port. E-LMI interoperates with an OAM
   protocol, such as Connectivity Fault Management (CFM), that runs
   within the provider network to collect OAM status. CFM runs at the
   provider maintenance level (UPE to UPE with inward-facing MEPs at
   the UNI). E-LMI relies on the OAM Ethernet Infrastructure (EI) to
   interwork with CFM for end-to-end status of Ethernet virtual
   connections (EVCs) across CFM domains.

   The LLDP frames using L2CP MAC DA of 01-80-C2-00-00-07 are peered or
   discarded by every node except for the Two Port MAC Relay (TPMR)
   bridge, so this kind of frame will not be overlaid across the L2
   overlay network.

    4.7. LLDP

   The Link Layer Discovery Protocol (LLDP) is a vendor-neutral link
   layer protocol in the Internet Protocol Suite used by network
   devices for advertising their identity, capabilities, and neighbors
   on an IEEE 802 local area network, principally wired Ethernet. The
   protocol is formally referred to by the IEEE as Station and Media
   Access Control Connectivity Discovery specified in standards
   document [IEEE 802.1AB].

   The LLDP frames using L2CP MAC DA of 01-80-C2-00-00-0E are peered or
   discarded by every node, so this kind of frame will not be overlaid
   across the L2 overlay network.

    4.8. PTP Peer Delay

   PTP Peer Delay frame is specified in [IEEE 1588-2008] to carry PTP
   peer time information. It uses L2CP MAC DA of 01-80-C2-00-00-0E and
   peered or discarded by every node, so this kind of frame will not be
   overlaid across the L2 overlay network.

    4.9. ESMC

   Ethernet Synchronization Messaging Channel (ESMC) is specified in
   [ITU-T Rec. G.8264] for conveying clock information between
   Synchronous Ethernet (SyncE) bridges.

Xia, et al.                                                 [Page 13]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   The ESMC frames using L2CP MAC DA of 01-80-C2-00-00-02 are peered or
   discarded by every node, so this kind of frame will not be overlaid
   across the L2 overlay network.

    4.10. GARP/MRP Block

   Multiple Registration Protocol (MRP), which replaced Generic
   Attribute Registration Protocol (GARP), is a generic registration
   framework defined by the [IEEE 802.1ak] amendment to the [IEEE
   802.1Q] standard. MRP allows bridges, switches or other similar
   devices to be able to register and de-register attribute values,
   such as VLAN identifiers and multicast group membership across a
   large LAN. MRP operates at the Data Link Layer.

   The block of L2CP MAC DA from 01-80-C2-00-00-20 to 01-80-C2-00-00-2F
   is used for MRP protocol. Now, only 01-80-C2-00-00-20 is for
   Multiple MAC Registration Protocol (MMRP) and 01-80-C2-00-00-21 is
   for Multiple VLAN Registration Protocol (MVRP), other L2CP MAC DA of
   the block are all reserved for future use. Protocol using one
   address of this block is passed by all the intervening bridges that
   does not participate in the protocol using this address, and peered
   or discarded by the bridge that participate in the protocol at last.
   In order to send the MRP frames to all related nodes (i.e., NVEs,
   bridges, etc) in one L2 overlay network, the MRP frames may require
   to be overlaid across the L2 overlay network.

5. L2CP Processing in L2GWs

   For all L2CP protocols, several differences exist between L2 overlay
   network and L2 bridge network on how to process them. As the
   demarcation point between L2 overlay network and L2 bridge network,
   L2GW keeps the same action to all L2CP frames as before at the L2
   bridge network side on the one hand, but maybe processes some L2CP
   frames differently at the L2 overlay network side on the other hand.
   The following sub sections will describe the L2CP process in L2GW.

   5.1. L2CP Frames Filtered (Peered or Discarded) in L2GW

   Although xSTP protocols using Nearest Customer Bridge address of 01-
   80-C2-00-00-00 indicate that it can be overlaid across L2 overlay
   network, they still are not necessary for L2 overlay network because
   L2 overlay network has its own mechanism to prevent bridge loops. So
   xSTP frames will be filtered by the L2GW and not go into the L2
   overlay network.

   Based on the analysis of section 3.3, LACP/LAMP frames using IEEE
   802 Slow Protocols of 01-80-C2-00-00-02 are not necessary for L2

Xia, et al.                                                 [Page 14]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   overlay network.  So, LACP/LAMP frames will be filtered by the L2GW
   and not go into the L2 overlay network. ESMC frames using the same
   MAC DA will also be filtered by L2GW.

   For Link OAM frames, if OAM functions are necessary for the whole L2
   network which interconnects L2 bridge network and L2 overlay network,
   L2GW needs to support the interworking of OAM as well. This means
   that L2GW should peer the Link OAM frames of L2 bridge network and
   perform some actions between NVEs in L2 overlay network. The
   detailed operation is TBD.

   Other L2CP protocols that are filtered by L2GW and do not go into L2
   overlay network include PAUSE, E-LMI, LLDP, PTP Peer Delay. The
   basic reason is that they all require to be processed hop by hop in
   L2 network strictly, but overlay network breaks this rule.

   The action of ''filter'' can be ''peer'', or ''discard''. It depends on
   the specific service requirement, i.e., does L2GW need to
   participate in the L2CP protocol, etc. How to determine the specific
   action is TBD.

   5.2. L2CP Frames Passed through L2GW

   Excepting for the aforementioned L2CP protocols filtered by L2GW,
   the left L2CP protocols need to be passed through L2GW. They include:

      LACP/LAMP frames using IEEE 802 Slow Protocols of 01-80-C2-00-00-

      GARP/MRP series protocols (i.e., MMRP, MVRP) using the MAC DA
      block of 01-80-C2-00-00-20 through 01-80-C2-00-00-2F.

   All these kinds of L2CP frames are passed through L2GW and traverse
   across the L2 overlay network and L2 bridge network to arrive the
   bridges that participate in the L2CP protocols. For MRP protocols,
   another necessary operation of L2GW is to use the pre-provisioned
   VLAN to virtual network instance (VNI) mappings in NVE locally or by
   getting from NVA to map these MRP frames into corresponding VNIs.

6. Other Interworking Cases

   There are other L2 bridge network technologies that use L2 Control
   Plane protocols such as Provider Bridge [IEEE802.1AD] or Provider
   Backbone Bridge [PBB] [IEEE802.1AH]. The use case of L2 Overlay

Xia, et al.                                                 [Page 15]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   Network interworking with these types of bridge networks is for the
   further study.

   Note that VPLS [RFC4761] [RFC4762], EVPN [EVPN], Shortest Path
   Bridging [IEEE SPB] and TRILL [RFC6325] are also technologies for L2
   private network implementation. These technologies rely on the
   control plane protocol and aim for service provider network. SDN
   controller interworking with such control plane protocol will be
   addressed in separate draft.

7. Security Considerations


8. IANA Considerations

   The document does not require any IANA action.

9. References

    9.1. Normative References

   [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
   Requirement Levels", BCP 14, RFC2119, March 1997.

   [RFC4761] Kompella, K. and Rekhter, Y. (Editors), "Virtual Private
   LAN Service (VPLS) Using BGP for Auto-Discovery and Signaling", RFC
   4761, January 2007

   [RFC4762] Lasserre, M. and Kompella, V. (Editors), "Virtual Private
   LAN Service (VPLS) Using Label Distribution Protocol (LDP)
   Signaling", RFC 4762, January 2007.

   [RFC6325]  Perlman, R., "RBridges: Base Protocol Specification",
   July 2011.

    9.2. Informative References

   [NVO3ARCH] Black, D, Narten, T., et al, "An Architecture for Overlay
   Networks (NVO3)", draft-narten-nvo3-arch-01, work in progress

   [NVO3FRWK] LASSERRE, M., Motin, T., et al, "Framework for DC Network
   Virtualization", draft-ietf-nvo3-framework-07, work in progress.

   [NVGRE]  Sridharan, M., et al, "NVGRE: Network Virtualization using
   Generic Routing Encapsulation", draft-sridharan-virtualization-
   nvgre-03, work in progress

Xia, et al.                                                 [Page 16]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   [VXLAN]  Mahalingam, M., Dutt, D., etc, "VXLAN: A Framework for
   Overlaying Virtualized Layer 2 Networks over Layer 3 Networks",
   draft-mahalingam-dutt-dcops-vxlan-05.txt, work in progress

   [EVPN] Sajassi, A. and R. Aggarwal, "BGP MPLS Based Ethernet VPN",
   draft-ietf-l2vpn-evpn-07, May 2014

   [EVPN-REQ] A. Sajassi, R. Aggarwal et. al., "Requirements for
   Ethernet VPN", RFC7209

   [EVPN-MHN] Weiguo, Hao, Yizhou, Li, et al, "Multi-homed network in
   EVPN", draft-hao-l2vpn-evpn-mhn-00, work in progress

   [802.1Q]   IEEE, "Media Access Control (MAC) Bridges and Virtual
   Bridged Local Area Networks", IEEE Std 802.1Q-2011, August, 2011.

   [IEEE 802.3-2005] "Part 3: Carrier sense multiple access with
   collision detection (CSMA/CD) access method and physical layer

   [IEEE 802.1AXbk-2012] "IEEE Standard for Local and metropolitan area
   networks--Link Aggregation Amendment 1: Protocol Addressing"

   [IEEE 802.3ah] "IEEE Standard for Information technology--Local and
   metropolitan area networks--Part 3: CSMA/CD Access Method and
   Physical Layer Specifications Amendment: Media Access Control
   Parameters, Physical Layers, and Management Parameters for
   Subscriber Access Networks"

   [IEEE 802.1X] "IEEE Standard for Local and metropolitan Area
   Networks. Port-based Network Access Control"

   [IEEE 802.1AB] "IEEE Standard for Station and Media Access Control,
   Connectivity Discovery"

   [MEF-16]  Metro Ethernet Forum, MEF 16, Ethernet Local Management
   Interface (E-LMI), January 2006.

   [IEEE 1588-2008] "IEEE Standard for a Precision Clock
   Synchronization Protocol for Networked Measurement and Control

   [IEEE 802.1ak] "IEEE Standard for Local and metropolitan Area
   Networks - Virtual Bridged Local Area Networks, Amendment 7:
   Multiple Registration Protocol"

Xia, et al.                                                 [Page 17]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015

   [IEEE 802.1AD], "Virtual Bridged Local Area Networks - Amendment 4:
   Provider Bridges", 2005

   [PBB] Clauses 25 and 26 of "IEEE Standard for Local and metropolitan
   area networks - Media Access Control (MAC) Bridges and Virtual
   Bridged Local Area Networks", IEEE Std 802.1Q, 2013.

   [IEEE802.1AH] IEEE Draft P802.1ah/D4.2 "Virtual Bridged Local Area
   Networks, Amendment 6: Provider Backbone Bridges", 2008

   [IEEE SPB] "IEEE standard for local and metropolitan area networks:
   Media access control (MAC) bridges and virtual bridged local area
   networks -- Amendment 20: Shortest path bridging", IEEE 802.1aq,
   June 2012.

   [ITU-T Rec. G.8264] "Distribution of Timing Through Packet Networks"

   Authors' Addresses

   Liang Xia (Frank)
   Huawei Technologies


   Lucy Yong
   Huawei Technologies, USA


   Weiguo Hao
   Huawei Technologies
   101 Software Avenue,
   Nanjing 210012

   Phone: +86-25-56623144

   Anoop Ghanwani

Xia, et al.                                                 [Page 18]

Internet-Draft          Layer 2 Gateway (L2GW)              April, 2015


   Ram (Ramki) Krishnan


Xia, et al.                                                 [Page 19]