E-Tree over MPLS: A Framework for Ethernet Tree (E-Tree) Service over a Multiprotocol Label Switching (MPLS) Network
How an MPLS network can offer a hub-and-spoke Ethernet service where the spokes cannot reach each other.
Key, R., Ed., Yong, L., Ed., Delord, S., Jounay, F., and L. Jin, "A Framework for Ethernet Tree (E-Tree) Service over a Multiprotocol Label Switching (MPLS) Network", RFC 7387, DOI 10.17487/RFC7387, October 2014, <https://www.rfc-editor.org/info/rfc7387>.
@misc{rfc7387,
series = {Request for Comments},
number = 7387,
howpublished = {RFC 7387},
publisher = {RFC Editor},
doi = {10.17487/RFC7387},
url = {https://www.rfc-editor.org/info/rfc7387},
author = {R. Key and L. Yong and S. Delord and F. Jounay and L. Jin},
title = {{A Framework for Ethernet Tree (E-Tree) Service over a Multiprotocol Label Switching (MPLS) Network}},
pagetotal = 13,
year = 2014,
month = oct,
} RFC text
Contents
Internet Engineering Task Force (IETF) R. Key, Ed.
Request for Comments: 7387
Category: Informational L. Yong, Ed.
ISSN: 2070-1721 Huawei
S. Delord
Telstra
F. Jounay
Orange CH
L. Jin
October 2014
A Framework for Ethernet Tree (E-Tree) Service
over a Multiprotocol Label Switching (MPLS) Network
Abstract
This document describes an Ethernet-Tree (E-Tree) solution framework
for supporting the Metro Ethernet Forum (MEF) E-Tree service over a
Multiprotocol Label Switching (MPLS) network. The objective is to
provide a simple and effective approach to emulate E-Tree services in
addition to Ethernet LAN (E-LAN) services on an existing MPLS
network.
Status of This Memo
This document is not an Internet Standards Track specification; it is
published for informational purposes.
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). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc7387.
Copyright Notice
Copyright (c) 2014 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
(http://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 Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction ....................................................3
1.1. Terminology ................................................3
2. Overview ........................................................4
2.1. Ethernet Bridge Network ....................................4
2.2. MEF Multipoint Ethernet Services: E-LAN and E-Tree .........4
2.3. IETF L2VPN .................................................5
2.3.1. Virtual Private LAN Service (VPLS) ..................5
2.3.2. Ethernet VPN (EVPN) .................................5
2.3.3. Virtual Private Multicast Service (VPMS) ............6
3. E-Tree Architecture Reference Model .............................6
4. E-Tree Use Cases ................................................8
5. L2VPN Gaps for Emulating MEF E-Tree Service .....................9
5.1. No Differentiation on AC Role ..............................9
5.2. No AC Role Indication or Advertisement .....................9
5.3. Other Issues ...............................................9
6. Security Considerations ........................................10
7. References .....................................................11
7.1. Normative References ......................................11
7.2. Informative References ....................................11
Acknowledgments ...................................................12
Contributors ......................................................12
Authors' Addresses ................................................13
1. Introduction
This document describes an Ethernet-Tree (E-Tree) solution framework
for supporting the Metro Ethernet Forum (MEF) E-Tree service over a
Multiprotocol Label Switching (MPLS) network. The objective is to
provide a simple and effective approach to emulate E-Tree services in
addition to Ethernet LAN (E-LAN) services on an existing MPLS
network.
This document extends the existing IETF-specified Layer 2 Virtual
Private Network (L2VPN) framework [RFC4664] to provide the emulation
of E-Tree services over an MPLS network. It specifies the E-Tree
architecture reference model and describes the corresponding
functional components. It also points out the gaps and required
extension areas in existing L2VPN solutions such as Virtual Private
LAN Service (VPLS) [RFC4761] [RFC4762] and Ethernet Virtual Private
Network (EVPN) [EVPN] for supporting E-Tree services.
1.1. Terminology
This document adopts all the terminologies defined in RFC 4664
[RFC4664], RFC 4761 [RFC4761], and RFC 4762 [RFC4762]. It also uses
the following terms:
Leaf Attachment Circuit (AC): An AC with Leaf role. An ingress
Ethernet frame at a Leaf AC (Ethernet frame arriving over an AC at
the Provider Edge (PE) of an MPLS network) can only be delivered
to one or more Root ACs in an E-Tree service instance. An ingress
Ethernet frame at a Leaf AC must not be delivered to any Leaf ACs
in the E-Tree service instance.
Root AC: An AC with Root role. An ingress Ethernet frame at a Root
AC can be delivered to one or more of the other ACs in the
associated E-Tree service instance.
E-Tree: An Ethernet VPN service in which each AC is assigned the role
of a Root or Leaf. The forwarding rules in an E-Tree are as
follows:
o The Root AC can communicate with other Root ACs and Leaf ACs.
o Leaf ACs can only communicate with Root ACs.
2. Overview
2.1. Ethernet Bridge Network
In this document, "Ethernet bridge network" refers to the Ethernet
bridge/switch network defined in IEEE 802.1Q [IEEE802.1Q]. In a
bridge network, a data frame is an Ethernet frame; data forwarding is
based on destination Media Access Control (MAC) address; MAC
reachability is learned in the data plane based on the source MAC
address and the port (or tagged port) on which the frame arrives; and
the MAC aging mechanism is used to remove inactive MAC addresses from
the MAC forwarding table on an Ethernet switch.
Data frames arriving at a switch may be destined to known unicast,
unknown unicast, multicast, or broadcast MAC destinations. Unknown
unicast, multicast, and broadcast frames are forwarded in a similar
way, i.e., to every port except the ingress port on which the frame
arrives. Multicast forwarding can be further constrained when using
multicast control protocol snooping or using multicast MAC
registration protocols [IEEE802.1Q].
An Ethernet host receiving an Ethernet frame checks the destination
address in the frame to decide whether it is the intended
destination.
2.2. MEF Multipoint Ethernet Services: E-LAN and E-Tree
MEF 6.1 [MEF6.1] defines two multipoint Ethernet Service types:
o E-LAN (Ethernet LAN), a multipoint-to-multipoint service
o E-Tree (Ethernet Tree), a rooted-multipoint service
The MEF defines User-Network Interface (UNI) in a multipoint service
as the Ethernet interface between Customer Equipment (CE) and a
Provider Edge (PE), i.e., the PE can send and receive Ethernet frames
to/from the CE. The MEF also defines UNI roles in a multipoint
service. One role is Root, and another is Leaf.
Note that MEF UNI in a service is equivalent to the Attachment
Circuit (AC) defined in L2VPN [RFC4664]. The Root AC and Leaf AC
defined in this document are the same as the Root UNI and Leaf UNI as
defined in MEF 10.3 [MEF10.3]. The terms "Root AC" and "Leaf AC" are
used in the following MEF service description.
For an E-LAN service, all ACs have the Root role, which means that
any AC can communicate with other ACs in the service. The E-LAN
service defined by the MEF may be implemented by IETF L2VPN solutions
such as VPLS and EVPN [EVPN].
An E-Tree service has one or more Root ACs and at least two Leaf ACs.
An E-Tree service supports communication among the roots and between
a root and a leaf but prohibits communication among the leaves.
Existing IETF L2VPN solutions can't support the E-Tree service. This
document specifies the E-Tree architecture reference model that
supports the E-Tree service defined by the MEF [MEF6.1]. Section 4
will discuss different E-Tree use cases.
2.3. IETF L2VPN
2.3.1. Virtual Private LAN Service (VPLS)
VPLS [RFC4761] [RFC4762] is an L2VPN solution that provides
multipoint-to-multipoint Ethernet connectivity across IP/MPLS
networks. VPLS emulates traditional Ethernet Virtual LAN (VLAN)
services in MPLS networks and may support MEF E-LAN services.
A data frame in VPLS is an Ethernet frame. Data forwarding in a VPLS
instance is based on the destination MAC address and the VLAN on
which the frame arrives. MAC reachability learning is performed in
the data plane based on the source address and the AC or pseudowire
(PW) on which the frame arrives. MAC aging is the mechanism used to
remove inactive MAC addresses from a VPLS switching instance (VSI) on
a PE. VPLS supports forwarding for known unicast frames, as well as
unknown unicast, broadcast, and multicast Ethernet frames.
Many service providers have deployed VPLS in their networks to
provide L2VPN services to customers.
2.3.2. Ethernet VPN (EVPN)
Ethernet VPN [EVPN] is an enhanced L2VPN solution that emulates an
Ethernet LAN or virtual LAN(s) across MPLS networks.
EVPN supports active-active multihoming of CEs and uses the
Multiprotocol Border Gateway Protocol (MP-BGP) control plane to
advertise MAC address reachability from an ingress PE to egress PEs.
Thus, a PE learns MAC addresses that are reachable over local ACs in
the data plane and other MAC addresses reachable across the MPLS
network over remote ACs via the EVPN MP-BGP control plane. As a
result, EVPN aims to support large-scale L2VPN with better resiliency
compared to VPLS.
EVPN is a relatively new technique and is still under development in
the IETF L2VPN WG.
2.3.3. Virtual Private Multicast Service (VPMS)
VPMS [VPMS] is an L2VPN solution that provides point-to-multipoint
connectivity across MPLS networks and supports various attachment
circuit (AC) types, including Frame Relay, ATM, Ethernet, PPP, etc.
In the case of Ethernet ACs, VPMS provides single coverage of
receiver membership, i.e., there is no differentiation among
multicast groups in one VPN. The destination address in the Ethernet
frame is not used in data forwarding.
VPMS supports unidirectional point-to-multipoint transport from a
sender to multiple receivers and may support reverse transport in a
point-to-point manner.
3. E-Tree Architecture Reference Model
Figure 1 illustrates the E-Tree architecture reference model. Three
Provider Edges -- PE1, PE2, and PE3 -- are shown in the figure. Each
PE has a Virtual Service Instance (VSI) associated with an E-Tree
service instance. A CE attaches to the VSI on a PE via an AC. Each
AC must be configured with a Root or Leaf role. In Figure 1, AC1,
AC2, AC5, AC6, AC9, and AC10 are Root ACs; AC3, AC4, AC7, AC8, AC11,
and AC12 are Leaf ACs. This implies that a PE (local or remote)
processes the Ethernet frames from CE01, CE02, etc., as if they
originated from a Root AC, and it processes the Ethernet frames from
CE03, CE04, etc., as if they originated from a Leaf AC.
Under this architecture model, the forwarding rules among the ACs,
regardless of whether the sending AC and receiving AC are on the same
PE or on different PEs, are described as follows:
o An egress frame (the frame to be transmitted over an AC) at an AC
with Root role must be the result of an ingress frame at an AC
(the frame received at an AC) that has Root or Leaf role and is
attached to the same E-Tree service instance.
o An egress frame at the AC with Leaf role must be the result of an
ingress frame at an AC that has Root role and is attached to the
same E-Tree service instance.
<------------E-Tree----------->
PE1+---------+ +---------+PE2
+----+ | +---+ | | +---+ | +----+
|CE01+----AC1----+--+ | | | | +--+----AC5----+CE05|
+----+ (Root AC) | | V | | | | V | | (Root AC) +----+
+----+ | | | | | | | | +----+
|CE02+----AC2----+--+ | | | | +--+----AC6----+CE06|
+----+ (Root AC) | | S +--+---------+--+ S | | (Root AC) +----+
+----+ | | | | | | | | +----+
|CE03+----AC3----+--+ | | | | +--+----AC7----+CE07|
+----+ (Leaf AC) | | I | | | | I | | (Leaf AC) +----+
+----+ | | | | | | | | +----+
|CE04+----AC4----+--+ | | | | +--+----AC8----+CE08|
+----+ (Leaf AC) | +-+-+ | | +-+-+ | (Leaf AC) +----+
+----+----+ +----+----+
| MPLS Core |
| +----+----+
| | +-+-+ | +----+
| | | +--+----AC9----+CE09|
| | | V | | (Root AC) +----+
| | | | | +----+
| | | +--+----AC10---+CE10|
+--------------+--+ S | | (Root AC) +----+
| | | | +----+
| | +--+----AC11---+CE11|
| | I | | (Leaf AC) +----+
| | | | +----+
| | +--+----AC12---+CE12|
| +---+ | (Leaf AC) +----+
PE3 +---------+
<-------------E-Tree---------->
Figure 1: E-Tree Architecture Reference Model
These rules apply to all frame types, i.e., known unicast, unknown
unicast, broadcast, and multicast. For known unicast frames,
forwarding in a VSI context is based on the destination MAC address.
A VSI on a PE corresponds to an E-Tree service instance and maintains
a MAC forwarding table that is isolated from other VSI tables on the
PE. It also keeps track of local AC roles. The VSI receives a frame
from an AC or across the MPLS core, and it forwards the frame to
another AC over which the destination is reachable according to the
VSI forwarding table and forwarding rules described above. When the
target AC is on a remote PE, the VSI forwards the frame to the remote
PE over the MPLS core. Forwarding over the MPLS core will be
dependent on the E-Tree solution. For instance, a solution may adopt
PWs to mesh VSIs as in VPLS and to forward frames over VSIs subject
to the E-Tree forwarding rules. Alternatively, a solution may adopt
the EVPN forwarding paradigm constrained by the E-Tree forwarding
rules. Thus, solutions that satisfy the E-Tree requirements could be
extensions to VPLS and EVPN.
In most use cases, an E-Tree service has only a few Root ACs (root CE
sites) but many Leaf ACs (leaf CE sites). Furthermore, a PE may have
only Root ACs or only Leaf ACs. Figure 1 provides a general E-Tree
architecture model.
4. E-Tree Use Cases
Table 1 below presents some major use cases for E-Tree.
+---------------------------+--------------+------------+
| Use Case | Root AC | Leaf AC |
+---+---------------------------+--------------+------------+
| 1 | Hub & Spoke VPN | Hub Site | Spoke Site |
+---+---------------------------+--------------+------------+
| 2 | Wholesale Access | Customer's | Customer's |
| | | Interconnect | Subscriber |
+---+---------------------------+--------------+------------+
| 3 | Mobile Backhaul | RAN NC | RAN BS |
+---+---------------------------+--------------+------------+
| 4 | IEEE 1588 PTPv2 [IEEE1588]| PTP Server | PTP Client |
| | Clock Synchronization | | |
+---+---------------------------+--------------+------------+
| 5 | Internet Access | BNG Router | Subscriber |
| | Reference [TR-101] | | |
+---+---------------------------+--------------+------------+
| 6 | Broadcast Video | Video Source | Subscriber |
| | (unidirectional only) | | |
+---+---------------------------+--------------+------------+
| 7 | Broadcast/Multicast Video | Video Source | Subscriber |
| | plus Control Channel | | |
+---+---------------------------+--------------+------------+
| 8 | Device Management | Management | Managed |
| | | System | Device |
+---+---------------------------+--------------+------------+
Where:
RAN: Radio Access Network NC: Network Controller
BS: Base Station PTP: Precision Time Protocol
BNG: Broadband Network Gateway
Table 1: E-Tree Use Cases
Common to all use cases, direct Layer 2 leaf-to-leaf communication is
required to be prohibited. For mobile backhaul, this may not be
valid for Long Term Evolution (LTE) X2 interfaces; an LTE X2
interface [LTE] enables communication between two evolved Node Bs
(eNBs). E-Tree service is appropriate for such use cases.
Also common to the use cases mentioned above, there may be one or
multiple Root ACs in one E-Tree service. The need for multiple Root
ACs may be driven by a redundancy requirement or by having multiple
serving sites. Whether a particular E-Tree service needs to support
one or multiple Root ACs depends on the application.
5. L2VPN Gaps for Emulating MEF E-Tree Service
The MEF E-Tree service defines special forwarding rules that prohibit
forwarding Ethernet frames among leaves. This poses some challenges
to IETF L2VPN solutions such as VPLS and EVPN in emulating E-Tree
service over an MPLS network. There are two major issues described
in the following subsections.
5.1. No Differentiation on AC Role
IP/MPLS L2VPN architecture has no distinct roles on Attachment
Circuits (ACs) and supports any-to-any connectivity among all ACs.
It does not have any mechanism to support forwarding constraints
based on an AC role. However, the MEF E-Tree service defines two AC
roles -- Root and Leaf -- and defines the forwarding rules based on
the originating and receiving AC roles of a given frame.
5.2. No AC Role Indication or Advertisement
In an L2VPN, when a PE, say PE2, receives a frame from another PE,
say PE1, over the MPLS core, PE2 does not know if the frame from PE1
is originated from a Root AC or Leaf AC. This causes the forwarding
issue on PE2 because the E-Tree forwarding rules require that the
forwarder must know the role of the frame origin, i.e., from Root AC
or Leaf AC. This is specifically important when PE2 has both Root AC
and Leaf AC attached to the VSI. E-Tree forwarding rules apply to
all types of frames (known unicast destination, unknown unicast
destination, multicast, and broadcast).
5.3. Other Issues
Some desirable requirements for IETF E-Tree are specific to an
IP/MPLS L2VPN implementation such as Leaf-only PE. A Leaf-only PE is
a PE that only has Leaf AC(s) in an E-Tree service instance; thus,
other PEs on the same E-Tree service instance do not necessarily
forward the frames originated from a Leaf AC to the Leaf-only PE,
which may save some network resources. It is also desirable for an
E-Tree solution to work with existing PEs that support single-role
ACs, where the role is equivalent to the root in an E-Tree service.
These requirements are described in the E-Tree requirement document
[RFC7152].
6. Security Considerations
An E-Tree service may be deployed for security reasons to prohibit
communication among sites (leaves). An E-Tree solution must enforce
E-Tree forwarding constraints. The solution must also guarantee that
Ethernet frames do not leak outside of the E-Tree service instance to
which they belong.
An E-Tree service prohibits communication among leaf sites but does
not have knowledge of higher-layer security constraints. Therefore,
in general, higher-layer applications cannot rely on E-Tree to
provide security protection unless all security constraints are fully
implemented by the E-Tree service.
Enhancing L2VPN for E-Tree services inherits the same security issues
described in the L2VPN framework document [RFC4664]. These relate to
both control-plane and data-plane security issues that may arise in
the following areas:
o issues fully contained in the provider network
o issues fully contained in the customer network
o issues in the customer-provider interface network
The framework document has substantial discussions on the security
issues and potential solutions to address them. An E-Tree solution
must consider these issues and address them properly. VPLS [RFC4761]
[RFC4762] and/or EVPN [EVPN] will likely be candidate solutions for
an E-Tree service over an MPLS network. The security capabilities
built into those solutions will be naturally adopted when supporting
E-Tree. For details, see the Security Considerations sections in
[RFC4761], [RFC4762], and [EVPN].
7. References
7.1. Normative References
[MEF6.1] Metro Ethernet Forum, "Ethernet Services Definitions -
Phase 2", MEF 6.1, April 2008.
[MEF10.3] Metro Ethernet Forum, "Ethernet Service Attributes -
Phase 3", MEF 10.3, October 2013.
[RFC4664] Andersson, L., Ed., and E. Rosen, Ed., "Framework for
Layer 2 Virtual Private Networks (L2VPNs)", RFC 4664,
September 2006,
<http://www.rfc-editor.org/info/rfc4664>.
[RFC4761] Kompella, K., Ed., and Y. Rekhter, Ed., "Virtual Private
LAN Service (VPLS) Using BGP for Auto-Discovery and
Signaling", RFC 4761, January 2007,
<http://www.rfc-editor.org/info/rfc4761>.
[RFC4762] Lasserre, M., Ed., and V. Kompella, Ed., "Virtual
Private LAN Service (VPLS) Using Label Distribution
Protocol (LDP) Signaling", RFC 4762, January 2007,
<http://www.rfc-editor.org/info/rfc4762>.
[RFC7152] Key, R., Ed., DeLord, S., Jounay, F., Huang, L., Liu,
Z., and M. Paul, "Requirements for Metro Ethernet Forum
(MEF) Ethernet-Tree (E-Tree) Support in Layer 2 Virtual
Private Network (L2VPN)", RFC 7152, March 2014,
<http://www.rfc-editor.org/info/rfc7152>.
7.2. Informative References
[IEEE802.1Q] IEEE, "IEEE Standard for Local and metropolitan area
networks -- Media Access Control (MAC) Bridges and
Virtual Bridged Local Area", IEEE Std 802.1Q, 2011.
[IEEE1588] IEEE, "IEEE Standard for a Precision Clock
Synchronization Protocol for Networked Measurement and
Control Systems", IEEE Std 1588, July 2008.
[LTE] 3GPP, "Evolved Universal Terrestrial Radio Access
(E-UTRA) and Evolved Universal Terrestrial Radio Access
Network (E-UTRAN)", 3GPP TS 36.300 v11.2.0, July 2012.
[TR-101] Broadband Forum, "Migration to Ethernet-Based Broadband
Aggregation", TR-101 Issue 2, July 2011.
[VPMS] Kamite, Y., Jounay, F., Niven-Jenkins, B., Brungard, D.,
and L. Jin, "Framework and Requirements for Virtual
Private Multicast Service (VPMS)", Work in Progress,
draft-ietf-l2vpn-vpms-frmwk-requirements-05,
October 2012.
[EVPN] Sajassi, A., Ed., Aggarwal, R., Bitar, N., Isaac, A.,
and J. Uttaro, "BGP MPLS Based Ethernet VPN", Work in
Progress, draft-ietf-l2vpn-evpn-11, October 2014.
Acknowledgments
The authors would like to thank Nabil Bitar and Adrian Farrel for
their detailed review and suggestions.
Contributors
The following people contributed significantly to this document.
Yuji Kamite
NTT Communications Corporation
Granpark Tower
3-4-1 Shibaura, Minato-ku
Tokyo 108-8118, Japan
EMail: y.kamite@ntt.com
Wim Henderickx
Alcatel-Lucent
Copernicuslaan 50
2018 Antwerp, Belgium
EMail: wim.henderickx@alcatel-lucent.com
Authors' Addresses
Raymond Key (editor)
EMail: raymond.key@ieee.org
Lucy Yong (editor)
Huawei USA
EMail: lucy.yong@huawei.com
Simon Delord
Telstra
EMail: simon.delord@gmail.com
Frederic Jounay
Orange CH
4 rue caudray 1020 Renens
Switzerland
EMail: frederic.jounay@orange.ch
Lizhong Jin
EMail: lizho.jin@gmail.com
ISO Plain Language
On this page
§1What the Document Sets Out to Do
The document describes a framework for the Metro Ethernet Forum E-Tree service over a Multiprotocol Label Switching network. The objective is a simple and effective approach to emulating E-Tree services alongside the ordinary Ethernet LAN service on a network that already exists. It extends the IETF Layer 2 VPN framework with a reference model and the functional components that go with it, and it points out the gaps in existing Layer 2 VPN solutions such as VPLS and Ethernet VPN that would have to be filled to support E-Tree.
The terminology follows RFC 4664, RFC 4761 and RFC 4762, with three additions. A Leaf attachment circuit is an attachment circuit with the Leaf role, meaning an Ethernet frame arriving at it can be delivered only to one or more Root attachment circuits in the same service instance, never to another Leaf. A Root attachment circuit is one with the Root role, meaning a frame arriving at it can be delivered to one or more of the other attachment circuits in the instance. An E-Tree is an Ethernet VPN service in which every attachment circuit has one of the two roles: a Root can talk to other Roots and to Leaves, and a Leaf can talk only to Roots.
The document explains how to give an MPLS network a new kind of Ethernet service, in which each customer site is marked either as a hub or as a spoke. The hub may talk to everybody; the spoke may talk only to hubs. The rest of the document is the model that makes this work and the list of what is still missing.
§2E-LAN, E-Tree and the Existing Toolkits
An Ethernet bridge network is the one defined in IEEE 802.1Q: a frame is forwarded on its destination MAC address, reachability is learned in the data plane from the source MAC address and the port the frame arrived on, and a MAC aging mechanism removes inactive addresses from the forwarding table. Frames may be for a known unicast address, or unknown unicast, multicast or broadcast, and the last three are flooded to every port except the one the frame arrived on.
The Metro Ethernet Forum defines two multipoint Ethernet service types in MEF 6.1: E-LAN, a multipoint-to-multipoint service, and E-Tree, a rooted multipoint service. The Forum also defines roles for the User-Network Interface, the Ethernet interface between customer equipment and the provider edge, which is the same thing as the attachment circuit of the Layer 2 VPN framework. For an E-LAN service every attachment circuit has the Root role, so any of them can talk to any other, and that is what VPLS and Ethernet VPN deliver. An E-Tree service has one or more Root attachment circuits and at least two Leaf ones. It allows communication among the roots and between a root and a leaf, and it prohibits communication among the leaves.
The existing tools differ in how they reach that result. VPLS provides multipoint-to-multipoint Ethernet across IP and MPLS networks, forwarding on the destination MAC address and the VLAN, learning reachability in the data plane from the source address and the attachment circuit or pseudowire the frame arrived on, and aging inactive addresses out of the switching instance. Many providers have deployed it. Ethernet VPN adds active-active multihoming of customer equipment and uses the Multiprotocol Border Gateway Protocol control plane to advertise MAC reachability from an ingress provider edge to egress ones, so a router learns local addresses in the data plane and remote ones from the control plane; it was still under development in the Layer 2 VPN working group when the document was written. Virtual Private Multicast Service offers point-to-multipoint connectivity with several attachment circuit types, and for Ethernet it gives every multicast group in a VPN the same coverage, ignoring the destination address in the frame.
Two services sit side by side. An E-LAN is a flat service where every site can talk to every other, and the existing Layer 2 VPN work already provides it. An E-Tree is the same kind of service with a hub at the top: the hubs can talk to each other and to the spokes, but the spokes cannot talk to each other. Nothing in the existing work can enforce that yet.
| Aspect | E-LAN | E-Tree |
|---|---|---|
| Attachment circuit roles | All circuits have the Root role | One or more Root circuits and at least two Leaf circuits |
| Who may talk to whom | Any circuit can talk to any other | Roots talk to roots and leaves; leaves talk only to roots |
| Leaf to leaf traffic | Allowed | Prohibited |
| Delivered today by | VPLS and Ethernet VPN | No existing Layer 2 VPN solution |
§3The E-Tree Reference Model
The reference model has three provider edges, each with a virtual service instance that corresponds to one E-Tree service. Customer equipment attaches to each provider edge over an attachment circuit, and every attachment circuit is configured with a Root or a Leaf role. In the example, six circuits are Root and six are Leaf. The rules that follow apply whether the sending and receiving circuits are on the same provider edge or on different ones.
An egress frame at a Root circuit must be the result of an ingress frame at a circuit with either role, attached to the same service instance. An egress frame at a Leaf circuit must be the result of an ingress frame at a circuit with the Root role. These rules apply to every frame type, known unicast, unknown unicast, broadcast and multicast, and for known unicast the forwarding within the instance is on the destination MAC address as usual.
Each virtual service instance keeps a MAC forwarding table that is separate from the other instances on the same provider edge, and it also records the role of its local circuits. A frame arriving from a circuit or across the MPLS core is forwarded to another circuit according to the table and the rules above, and when the target circuit is on another provider edge the instance sends it over the MPLS core. How that happens depends on the solution: one option is to mesh the instances with pseudowires as VPLS does and apply the E-Tree rules when forwarding, and another is to use the Ethernet VPN forwarding paradigm constrained by the E-Tree rules. Solutions that meet the E-Tree requirements would therefore be extensions to VPLS and to Ethernet VPN. In most uses an E-Tree has a few Root circuits, at the hub sites, and many Leaf circuits, and a provider edge may carry only Roots or only Leaves.
The model is simple: mark each customer connection as a hub or a spoke, and then apply two rules when forwarding a frame. A frame leaving a hub may have come from anywhere. A frame leaving a spoke must have come from a hub. The service instance keeps its own table of which MAC addresses are reachable where, so the two rules are enforced inside the normal Ethernet forwarding.
- E-Tree service instance
- PE1
- AC1 Root CE01
- AC2 Root CE02
- AC3 Leaf CE03
- AC4 Leaf CE04
- PE2
- AC5 Root CE05
- AC6 Root CE06
- AC7 Leaf CE07
- AC8 Leaf CE08
- PE3
- AC9 Root CE09
- AC10 Root CE10
- AC11 Leaf CE11
- AC12 Leaf CE12
- PE1
§4Where E-Tree Is Used
The document lists eight uses. A hub and spoke VPN has the hub site as the Root and the spoke sites as Leaves. Wholesale access has the customer's interconnect as the Root and the customer's subscribers as Leaves. Mobile backhaul has the radio access network controller as the Root and the base stations as Leaves. Precision Time Protocol clock synchronisation has the time server as the Root and the clients as Leaves. An internet access reference has the broadband network gateway router as the Root and the subscribers as Leaves. Broadcast video, which is unidirectional only, has the video source as the Root and the subscribers as Leaves. Broadcast or multicast video with a control channel has the video source as the Root and the subscribers as Leaves. Device management has the management system as the Root and the managed devices as Leaves.
Common to all of them is that direct communication between Leaf sites has to be prohibited. For mobile backhaul the document notes an exception, since an LTE X2 interface allows communication between two base stations, and an E-Tree is appropriate for such a use. Also common is that a service may have one or several Root circuits, several being driven by a redundancy requirement or by having more than one serving site, and which applies depends on the application.
The document lists the situations where a one-to-many service is the right shape: a company's hub with branch offices, a provider giving wholesale access to its own customers, mobile base stations talking to a controller, time servers feeding clocks, a broadband gateway feeding subscribers, a video source feeding viewers, and a management system talking to the devices it manages. In all of them the many end sites must not be able to talk to each other.
- Hub and spoke VPNRoot: the hub site. Leaf: the spoke sites.
- Wholesale accessRoot: the customer interconnect. Leaf: the customer subscribers.
- Mobile backhaulRoot: the radio access network controller. Leaf: the base stations.
- Clock synchronisationRoot: the time server. Leaf: the clients.
- Internet accessRoot: the broadband network gateway router. Leaf: the subscribers.
- Broadcast videoRoot: the video source. Leaf: the subscribers.
- Video with control channelRoot: the video source. Leaf: the subscribers.
- Device managementRoot: the management system. Leaf: the managed devices.
§5The Gaps in Layer 2 VPN
Two problems stand in the way of emulating E-Tree with the existing work. The first is that the IP and MPLS Layer 2 VPN architecture has no distinct roles on attachment circuits and supports any-to-any connectivity between all of them, with no mechanism for forwarding constraints based on a role. The Metro Ethernet Forum service, by contrast, defines exactly two roles and derives its forwarding rules from the roles of the sending and receiving circuits.
The second is that no role is indicated or advertised. When a provider edge receives a frame from another provider edge over the MPLS core, it does not know whether the frame came from a Root circuit or a Leaf circuit, and the forwarding rules need that knowledge. The problem is sharpest when the receiving provider edge has both Root and Leaf circuits attached to the same service instance. It applies to every frame type, known unicast, unknown unicast, multicast and broadcast.
Some further requirements are specific to an IP and MPLS implementation. A Leaf-only provider edge is one that has only Leaf circuits in the service instance, and other provider edges need not forward frames that originated from a Leaf circuit to it, which may save network resources. It is also desirable for a solution to work with existing provider edges that support only a single role, where that role is the equivalent of the root. These requirements are set out in the E-Tree requirements document, RFC 7152.
Two things are missing from the current Layer 2 VPN work. Attachment circuits have no role at all, so there is nowhere to record that a site is a spoke. And when a frame crosses the network, the router at the far end is not told whether the frame came from a hub or from a spoke, so it cannot apply the rule. The document also notes that a provider edge serving only spokes need not receive frames from other spokes at all.
§5Security Considerations
An E-Tree service may be deployed for security reasons, to prohibit communication between sites, so a solution must enforce the E-Tree forwarding constraints. It must also guarantee that Ethernet frames do not leak outside the service instance to which they belong.
An E-Tree service prohibits communication among leaf sites but has no knowledge of the security constraints of higher layers, so in general higher-layer applications cannot rely on E-Tree for protection unless all the relevant constraints are implemented by the service itself.
Enhancing Layer 2 VPN for E-Tree inherits the security issues described in the Layer 2 VPN framework document, both in the control plane and in the data plane: issues entirely inside the provider network, issues entirely inside the customer network, and issues in the network between customer and provider. A solution must consider these and address them properly. VPLS and Ethernet VPN are the likely candidate solutions, and the security capabilities built into them are adopted along with them.
The service may exist precisely to keep sites apart, so the forwarding rules have to be enforced and frames must not leak into another service. But the service knows nothing about what the applications running over it need, so an application that relies on the service for protection must implement its own rules as well. Everything else follows the security rules of the Layer 2 VPN work it is built on.