next generation mobile networks
A White Paper by the NGMN Alliance
LTE Backhauling Deployment Scenarios
1
ngmn
LTE backhauling deployment scenarios
by NGMN Alliance
Version: 1.4.2 FINAL
Date: 3rd July 2011
Document Type: Final Deliverable (approved)
Confidentiality Class: P
Authorised Recipients: N/A
Project: P-OSB: Optimized Backhaul
Editor / Submitter: Miguel Angel Alvarez, Frederic Jounay, Tamas Major, Paolo Volpato
Contributors: NGMN Optimised Backhaul Project Group
Approved by / Date: BOARD July 3, 2011




This document contains information that is confidential and proprietary to NGMN Ltd. The information may not be
used, disclosed or reproduced without the prior written authorisation of NGMN Ltd., and those so authorised may
only use this information for the purpose consistent with the authorisation.


ⓒ 2011 Next Generation Mobile Networks Ltd. All rights reserved. No part of this document may be reproduced or
transmitted in any form or by any means without prior written permission from NGMN Ltd.
The information contained in this document represents the current view held by NGMN Ltd. on the issues
discussed as of the date of publication. This document is provided “as is” with no warranties whatsoever including
any warranty of merchantability, non-infringement, or fitness for any particular purpose. All liability (including liability
for infringement of any property rights) relating to the use of information in this document is disclaimed. No license,
express or implied, to any intellectual property rights are granted herein. This document is distributed for
informational purposes only and is subject to change without notice. Readers should not design products based on
this document.
2
Abstract
With the introduction of LTE operators need to look at how the backhauling network, the network domain
that connects evolved NodeBs (eNBs) to MME and S/P-GW, is capable of adapting to the new
requirements, namely the adoption of a packet infrastructure, without disrupting the existing services.
This paper introduces some reference architectures, moving from a pure layer 2 topology to a full layer 3
one, discussing some elements to be considered in the design process of a network. The purpose of this
is to support operators in their migration from current architectures to new, packet-based backhaul
networks.
Since the migration phase might pose concerns to operators still engaged in 2G/3G deployment or in
maximizing the profitability of their existing networks, the scenarios described hereafter have been
designed considering, among the other aspects, possible paths to migrate from circuit based network,
and represent a kind of target architecture to aim at. Clearly, the path to an all packet-based backhaul will
depend on the specific market, technical environment and services an operator is operating or dealing
with.
Although this paper primarily focuses on LTE backhauling, some advanced topics are also referenced
and constitute the basis for further studies on LTE-A.
3
Table of Contents
1. Introduction ........................................................................................................................... 4
1.1. NGMN TWG P11 - P-OSB Scope .................................................................................. 4
1.2. What this paper is and why it is in line with P-OBS ....................................................... 4
2. Definitions and abbreviations ................................................................................................ 4
3. Framework and terminology .................................................................................................. 5
3.1. HSPA+/LTE/4G requirements for backhauling .............................................................. 6
3.2. Evolving from current backhauling: main issues ............................................................ 7
3.3. Gluing the organic view and the logical view ................................................................. 9
3.4. Guidelines followed to define a logical scenario .......................................................... 11
4. RAN Configuration .............................................................................................................. 11
4.1. MPLS in the eNB ......................................................................................................... 12
5. LTE-ready backhauling scenarios ....................................................................................... 13
5.1. Technology .................................................................................................................. 13
5.2. Scenario 1 – Carrier Ethernet ...................................................................................... 14
5.2.1. Applicability ........................................................................................................... 14
5.2.2. Considered implementations (protocol stacks) ..................................................... 15
5.3. Scenario 2 – Carrier Ethernet + L2/L3 VPN ................................................................. 17
5.3.1. Applicability ........................................................................................................... 17
5.3.2. Considered implementations (protocol stacks) ..................................................... 18
5.4. Scenario 3 – MPLS access + L2/L3 VPN .................................................................... 19
5.4.1. Applicability ........................................................................................................... 20
5.4.2. Considered implementations (protocol stacks) ..................................................... 20
5.5. Scenario 4 – L2/L3 VPN in access + L2/L3 VPN in aggregation ................................. 22
5.5.1. Applicability ........................................................................................................... 22
5.5.2. Considered implementations (protocol stacks) ..................................................... 23
5.6. Scenario 5 – End-to-end (Multi-segment) Pseudowire ................................................ 26
5.6.1. Applicability ........................................................................................................... 26
5.6.2. Considered implementations (protocol stacks) ..................................................... 27
5.7. Scenario 6 – Full L3 ..................................................................................................... 28
5.7.1. Applicability ........................................................................................................... 28
5.7.2. Considered implementations (protocol stacks) ..................................................... 29
6. How to “pick” the right scenario ........................................................................................... 30
7. Open points ......................................................................................................................... 31
7.1. LTE-A ........................................................................................................................... 31
7.2. Security ........................................................................................................................ 31
7.3. Selective IP Offloading (SIPTO) .................................................................................. 32
7.4. Femto-Cells or none-3GPP Access ............................................................................. 32
8. Conclusions and next steps ................................................................................................ 32
9. References .......................................................................................................................... 33
4
1. Introduction
The goal of this paper is to describe how the mobile backhauling network will evolve to support the
deployment of LTE.
1.1. NGMN TWG P11 - P-OSB Scope
The Optimized Solutions for Backhaul and Meshed Networks (P-OSB) project aims to define
requirements and to assess innovative all-IP transport solutions for facilitating optimum backhauling
(including self-backhauling).
1.2. What this paper is and why it is in line with P-OBS
This paper addresses the logical topologies that can be deployed on top of the connectivity scenarios
described in [2].
2. Definitions and abbreviations
Abis Logical interface between 2G BTS
and BSC PDH Plesiochronous Digital Hierarchy
ATM Asynchronous Transfer Mode POS Packet Over Sonet
BGP Border Gateway Protocol PPP Point to Point Protocol
CAC Call Admission Control QoS Quality of Service
CE Customer Edge S1 Logical interface between LTE
BTS and packet core
CPE Customer Premises Equipment SDH Synchronous Digital Hierarchy
CSG Cell Site Gateway SGSN Serving GPRS Support Nodes
DSCP Differentiated Service Code Point TDM Time division Multiplex
EPC Evolved Packet Core TE Traffic Engineering
GGSN Gateway GPRS Support Node UMTS Universal Mobile
Telecommunication Service
GPRS General Packet Radio Service VC Virtual Circuit
GW Gateway VLAN Virtual LAN
HSPA High Speed Packet Access VPLS Virtual Private LAN Service
Iub Logical interface between 3G BTS
and RNC VPN Virtual Private Network
LSP Label Switched Path VRF Virtual Routing and Forwarding
LTE Long Term Evolution VSI Virtual Switching Instance
MPLS Multi Protocol Label Switching X2 Logical interface between LTE
BTS
MASG Mobile Aggregation Site Gateway
OSPF Open Shortest Path First
P Provider (Router)
PE Provider Edge (Router)
PHB Per Hop Behaviour
5
3. Framework and terminology
A backhaul network serves as the transport medium for a mobile Radio Access Network (RAN) and
connects the base stations to their relevant controllers. The term “controller” is used in the context of this
document as a representation of the complete EPC (Evolved Packet Core) including MME and S/P-GW in
case of LTE, and the controllers of other radio technologies like RNC in case of 3G and BSC in case of
2G.
In essence, a typical network built for mobile backhaul consists of three domains: Core, Aggregation and
Access. The domain borders are mostly defined by the technology and topology used within the domain
and the deployed radio nodes.
The access network provides the connectivity to the BTS at the cell sites and is predominantly based on
tree and chain topologies built with microwave radios, but also with a good share of fiber and copper
usage.
Starting from the aggregation network we see very often ring and mesh topologies, mostly on top of an
optical network flavor. The aggregation network is normally terminated at the controller site, where RNCs
and BSCs are mostly located.
The controller sites are connected to other controller sites and the packet core as well as the EPC
(Evolved Packet Core) via the core network, which is nearly in all cases an IP/MPLS routed network.
This structure is represented in the following picture.
First Core
Base mile Aggregation
station
Second Controller
mile
Service relationship (End-to-end)
Packet
node
Demarcation node
(CSG)
Physical
connectivity
Access Aggregation
Physical
connectivity
Physical
connectivity
Physical
connectivity
Physical
connectivity
Packet
node
Demarcation
node
(MASG)
Controller
Core
Networking layer(s)
Figure 1: Basic Structure of a Mobile Backhaul Network
For the scope of this paper, the backhaul network will be limited to only two network domains: access and
aggregation.
The reason why access may be composed of several sub-domains (first mile, second mile, etc.) is to
consider different physical technologies and topologies. Moving left to right, the first mile (first hop)
connects a demarcation device, usually deployed at the cell site, to a first stage of traffic grooming and
concentration. The second mile, in turn, further aggregates traffic, adapts any technology change and
provides the hand-over point to a metro/aggregation network.
6
Another demarcation device exists at the right border of the aggregation domain, connecting the backhaul
network directly to a RAN controller or to the network core.
The list of nodes that are part of a backhaul network then includes:
 a Cell Site Gateway (CSG, also referred to as Cell Site Aggregator, or Demarcation device),
usually deployed at a cell site, which is the first network node where the logical architectures
described hereafter apply
 some packet nodes belonging either to access or aggregation
 a Mobile Aggregation Site Gateway (MASG) which acts as a counterpart of the CSG.
The term “physical connectivity” is used to represent whatever technology can be used to connect nodes,
as explained later on.
On top of the physical layer a networking layer can be found. This is the focus of the present analysis;
again the term is wide enough to embrace all of the possible logical architectures needed to steer LTE
traffic and applications.
The highest layer is represented by the service, where this applies to the S1 and X2 interfaces. Even if
not explicitly mentioned, the transport of S1 and X2 relies on an IP stratum which is not part of
networking.
3.1. HSPA+/LTE/4G requirements for backhauling
The technical requirements of a backhaul network have been derived either from 3GPP specifications,
the primary reference for HSPA+ and LTE, and other industry/normative bodies’ specifications (IETF,
Metro Ethernet Forum, DSL Forum, ITU, etc.).
At a very high level, the basic requirement on a backhaul network is to support LTE, HSPA+, and in
general 4G transport, but for the scope of this analysis it can be summarized into the following points,
which are not necessarily always supported by current networks:
 The backhaul network is packet based
 Provides high bandwidth
 Network nodes are characterized by high capacity interfaces and perform QoS aware traffic
aggregation
 Enables end-to-end Operation, Administration & Maintenance (OAM)
 Possibly has a lower TCO (total cost of ownership) than traditional TDM or hybrid (TDM and
Ethernet) networks
 Supports the networking models and transport services as defined by MEF, BBF, IETF and any
other relevant industry organizations.
The majority of the LTE backhaul traffic follows still the traditional hub-and-spoke architecture, i.e. eNBs
send their S1 traffic towards the core network via a common peering point.
LTE introduces a new logical interface for BTS to BTS communication called X2, which was not existing
(and also not needed) with earlier 2G and 3G systems. Its main usage is in supporting the handover
process when a terminal is changing from one BTS to another.
The logical connectivity for the X2 traffic can be provided at various points in the backhaul network
7
At a first glance, it seems obvious that it would be beneficial if the X2 “turning point” would be located
close to the base stations, allowing to benefit from low X2 latency and avoiding that X2 traffic will load
higher parts of the Mobile Backhaul Network.
However keeping the X2 latency significantly lower than the radio link interruption time of 30 .. 50 ms
during handover does not add any value. As in a well designed network the S1 transport path is anyhow
optimized for low latency, it mostly does not cause any harm to provide the X2 connectivity at a higher
point in the network. Furthermore, the amount of X2 traffic is marginal compared to S1 traffic, so the
additional load is negligible.
3.2. Evolving from current backhauling: main issues
The introduction of LTE might pose some concerns to operators still engaged in 2G/3G deployment or in
maximizing the profitability of their existing networks. For this reason, the scenarios described hereafter
have been designed considering, among the other aspects, the shift from a more “traditional” network
(e.g. circuit based) and represent a kind of target architecture to aim at. Clearly, the path to an all-packet
based backhauling will depend on the specific market, technical environment and services an operator is
operating or dealing with.
Without the aim of being exhaustive, current backhaul networks supporting 2G/3G services are mostly
based on legacy technologies (TDM/ATM). In general, 2G Abis is carried on TDM, while 3G Iub transport
relies on ATM over TDM. Physical technologies may include PDH and/or SDH.
One first approach to tackle the move to a packet backhaul is the realization of a dedicated network for
HSPA+/LTE, whilst maintaining the existing network for current services. The parallel networks’ approach
leaves an operator the flexibility of gradually introducing packet-based technologies in certain segments
of the networks and selecting, when needed, the preferred method to handle legacy services. For
example, 2G/3G data services can be carried in their native form or can be offloaded to the new Ethernet
network (e.g. for HSDPA).
In selecting such an approach, scenarios based on Ethernet (please see 5.1) are likely to simplify the
interconnection of existing network domains and support the typical hub & spoke logical topology
between RAN controllers or EPC components and base stations. A high level schematic is represented
into the next picture.
Access Aggregation
Figure 2: High-level representation of a circuit-based network
8
Access Aggregation
A different approach might consider the adoption of Ethernet as the common backhaul technology, from
the cell site to the aggregation, used across only one network. The main impact is the upgrade to
Ethernet on every base station site and network element, even if Ethernet and legacy transport can
coexist in the access domain (hybrid approach). Legacy services are mapped onto the Ethernet layer
using pseudowires or circuit emulation techniques and encapsulated into e.g. VLANs. The aggregation
domain could still employ SDH, bringing an Ethernet over SDH transport. The typical topology is still hub
& spoke, as seen into the next picture, and reference scenarios could be 5.1 or 5.3.
Access Aggregation
Figure 3: High-level representation of a mixed circuit-based and packet-based network
One last approach foresees the deployment of an MPLS-based VPN spanning from cell sites to RAN
controllers or EPC components, either at layer 2 or 3. Ethernet may remain the underlying transport
technology but other solutions are possible (e.g. SDH, WDM for the aggregation, GPON, dark fiber,
microwave for the access).
Cell site gateway equipment is distributed at the cell sites and services are encapsulated as pseudowire
or circuit emulation. The supported topologies include hub & spoke, mesh and ring.
The overall architecture is represented in the next picture and can be considered by scenarios from 5.4
onwards.
Figure 4: High-level representation of a packet-based network
9
As said, the approaches can be seen as different alternatives or different steps in the move to packetbased
backhauling. Specifically the last one is often seen as the last to arrive on the market and to be
adopted by carriers.
3.3. Gluing the organic view and the logical view
Backhauling has been studied from a physical standpoint in [1]. The main technologies included in that
analysis are listed here for reference:
 Microwave point-to-point
 Microwave point-to-multipoint
 DSL
 GPON
 Ethernet Leased Line
 Fiber point-to-point
 Ring or mesh - Ethernet, NG-SDH, (D)WDM.
Depending on the physical transmission technologies Service Providers have deployed in the field and
how they are combined, a few topologies are possible for a backhaul network. This analysis has focused
on the following cases, considered as general combinations of access and aggregation topologies:
Case Access Aggregation Examples
1 Tree No aggregation Point-to-point or point-to-multipoint fiber
or microwave connections groomed by a
packet node in front of RAN controllers
2 Tree Mesh/Ring Point-to-point or point-to-multipoint fiber
or microwave connections with
aggregation by a metro Ethernet or SDH
network
3 Ring Mesh/Ring Fiber or microwave based access rings,
with metro Ethernet or SDH aggregation
4 Mesh Mesh/Ring Fiber or microwave based mesh in
access, with metro Ethernet or SDH
aggregation
10
To better visualize how a backhaul network is shaped the next table provides a few examples of
topologies.
Case Example 1 Example 2
1 Access Aggregation Access Aggregation
2 Access Aggregation Access Aggregation
3 Access Aggregation Access Aggregation
4 Access Aggregation
Access Aggregation
Legenda: Packet node Backhaul link eNB site Controller
11
Although high level, the table gives a first hint on the characteristics of topologies. These characteristics
will also be considered later on to determine what logical architectures can be applied to every case:
 Case 1 is characterized by direct connectivity with no or little path diversity for redundancy.
Protection tends to be at the transmission layer, posing less stringent requirements on the logical
architecture to be adopted;
 Case 2 limits the topology described by case 1 to the access, while a ring or mesh infrastructure
is considered in the aggregation, for a denser grooming of traffic. The same consideration made
for case 1 applies to the access, while the presence of a ring or even a mesh in the aggregation
suggests considering architectures supporting fast detection and reaction time;
 Case 3 and case 4 further increase the complexity of backhaul networks: the difference between
the two is how the access domain is shaped (ring for case 3 and mesh for case 4, where a ring is
seen as collapsed mesh). Depending of Service Providers’ preference, several architectures are
possible, all of them generally based on MPLS.
Please note that the nodes at the border between two domains (and providing the connection between
them), e.g.: between access and aggregation, are typically configured in a redundant manner for high
availability.
3.4. Guidelines followed to define a logical scenario
The physical and topological aspects of access and aggregation described earlier can pose some
requirements on the definition of logical architectures.
At the same time, other factors may determine an impact on that. Some of them have been already
mentioned in the previous paragraph, some depend on local preferences or guidelines. In general we
have:
 LTE traffic flows steering (from eNBs to MME and S/P-GW vs. from eNBs to eNBs)
 Underlying technology
 Need for physical or logical protection
 Constraints from the environment (e.g. radio based transport vs. fiber).
Then the following criteria have been identified for mapping topologies to functional architectures:
 Forwarding technology
 Control plane usage for protection
 Support of end-to-end OAM
 Preference of availability of end-to-end services (MEF based, L2VPN, L3VPN, etc.).
4. RAN Configuration
Most base stations support a flexible way to bind eNB applications (S1/X2 U-plane, S1/X2 C-plane, Mplane,
S-plane) arbitrarily to either
 eNB interface address(es) or
 eNB virtual (loopback) address(es).
This flexibility allows base stations to be configured according to the transport services offered by the
backhaul network, but also applying traffic separation (e.g. M-plane from U/C-plane traffic) as needed.
eNB interface IP address(es) can be assigned either to
 one or more physical interface(s) or
 one or more logical interface(s).
12
A physical interface is typically provided by an Ethernet port, whereas a logical interface is provided by a
VLAN termination. Different interfaces as well as VLANs belong to different IP subnets.
There are many configurations possible, but for the sake of simplicity only two exemplary configurations
will be shown:
Figure 5: Example IP configurations of eNB
In the left example one IP address is used for terminating control, user and synchronization plane traffic
and a second one solely for management plane traffic. As both IP addresses are running over different
logical interfaces created by two VLANs, the addresses have to belong to different IP subnets.
In the right example separate logical interfaces (VLAN) and IP addresses are used for the management
and synchronization plane. User and control plane traffic are terminated on individual virtual (also called
loopback) addresses, which can be reached via a dedicated transport IP address. In most cases such
configurations in the eNB are created by using BTS-internal routing functionality.
For both examples it has been assumed that IEEE1588 is used for synchronization purposes, but if this is
not the case the S-plane termination point becomes obsolete. Also only a single Ethernet interface on the
eNB has been considered, but also configurations with multiple Ethernet interfaces are possible.
In case IPSec is used then the configurations are typically done by using IPSec Tunnel mode. The IP
address for IPsec tunnel termination (interface IP address) may be different from the IP addresses which
eNB U/C/M/S-plane applications are bound to (virtual IP addresses). Configuration and usage in the
network would be similar to the case depicted in the right side of Figure 5.
Also other configurations can be foreseen: If for example the backhaul network is offering different
transport services for meshed X2 and hub-and-spoke S1 traffic, then a configuration with S1 U/C plane
on one logical interface and X2 U/C plane on another logical interface would be very reasonable.
4.1. MPLS in the eNB
Some scenarios described in chapter 5 of this document are based on MPLS and/or MPLS-TP in the
backhaul network, and in some scenarios even down to the cell site. The question coming immediately to
mind is whether it would make sense to extend in those cases the MPLS layer into the eNB itself.
From a concept perspective, this would be the equivalent of moving the demarcation device at the cell
site, also called Cell Site Gateway (CSG), into the eNB itself. The logical interfaces, as described in the
previous chapter, would be e.g. IP carried over an MPLS LSP.
3GPP has specified the S1 and X2 protocols to run over IP, but without being specific on the data link
layer technology to be used. They have indicated that suitable options are Ethernet and PPP.
13
Having this in mind, the usage of another networking layer below IP in the eNB – like MPLS – would not
be in contradiction to 3GPP standards.
The benefit of such an integration of MPLS into the eNB would be mainly a simplification of the network
architecture and a reduction of the required footprint at the cell site by making the CSG obsolete. It would
also help to improve the operation of the network as it would allow direct access to the MPLS OAM
functionality as well as the performance measurements within the BTS.
On the other hand there are certain disadvantages, which should be also considered: By basically moving
the CSG into the eNB, the interface between radio and transport equipment would turn into an NNI,
hence there would be no clear demarcation anymore between the radio and transport domain. Also the
operational complexity will be increased in case there is a separation between transport and radio
department within the operator’s organization. The main concern with such integration would be on the
interoperability between the eNB and the transport equipment, which would become more complex than
with a plain, simple and well defined IP/Ethernet interface.
We can conclude that there are several reasons which speak against and in favor of an integration of
MPLS into the eNB, and the operator has to decide whether the benefit of the footprint reduction at the
cell site is worth the extra effort.
5. LTE-ready backhauling scenarios
This chapter introduces a few scenarios that have been recognized as the most probable or interesting
ones during the analysis activity. They comprise a L2-oriented architecture up to a full L3 one, with
several steps in between.
5.1. Technology
For the sake of simplicity all scenarios have been defined only on top of Ethernet (IEEE 802.3), as it is
expected to be the dominant transport technology in future. Other technologies can be considered as
well, and at the end of section 5.2.2 an example with DWDM is given. The use of Ethernet interfaces
has been also assumed for all base station and controller types. With LTE this is anyhow the only defined
transport interfaces and more and more 3G as well as 2G systems are moving towards Ethernet
connectivity.
Another point worth mentioning is the way MPLS-TP and IP/MPLS are considered. Both have their own
applicability in the scenarios presented hereafter, each with its own specificity.
Whilst, in theory, a lot of protocol combinations are possible, from a practical point of view IP/MPLS can
be used for implementing both L2 and L3 VPNs. Often they can be found in the aggregation domain but
nothing prevents their adoption even in the access.
MPLS-TP has its applicability for point-to-point transport services, suitable in access and for those cases
where a point-to-point L2 VPN (VPWS) can be extended also into the aggregation.
14
5.2. Scenario 1 – Carrier Ethernet
The first scenario is characterized by Ethernet as the service layer that carries S1 and X2 traffic on top of
any transport network used either in access and aggregation. Scenario 1 is represented in the next
picture.
First
mile
Second
mile Aggregation Core
Service/
Networking
plane Controller
Ethernet Ethernet Ethernet / DWDM Ethernet
IP
Ethernet
S1
MEF EVC (VLAN-based Service Layer, e.g. E-Line or E-LAN)
Demarcation
node
Packet
node
Packet
node
Demarcation
node
Access Aggregation
any network
X2
Core
eNB
Figure 6: Backhaul Scenario based on Carrier Ethernet
The traffic steering is based upon VLAN, defining Ethernet Virtual Connections between eNBs and MME
or S/P-GW. The service layer might rely on any of MEF’s models (E-Line, E-LAN, E-Tree), as
appropriate.1
While there is no doubt the S1 interface spans from an eNB to EPC components, the X2 interface can be
switched at any of the packet nodes in access or aggregation. A Service Provider can choose whether to
favor a lower latency (retaining the switching point close to the eNBs) or a tighter control of the traffic
(thus moving the X2 switching point close to the EPC).
5.2.1. Applicability
This scenario does not rely on IP/MPLS for redundancy and protection. Reliability is provided by
protection mechanisms at the transmission (physical) layer (e.g. microwave 1+1 hot-standby, leased lines
in LAG, etc.). Even if not exclusive to that, this scenario is considered to fit into case 1.
End-to-end OAM is at Ethernet level (802.1ag, 802.3ah, Y.1731 are examples of available tools).

1 For harmonization with the protocol stacks figures as well as with for those using different transport technologies,
the service termination point has been placed in the middle of the demarcation node and by that in contradiction to
MEF specifications. According to MEF terminology the UNI is actually between e.g. the demarcation node (CSG) and
the eNB

15
5.2.2. Considered implementations (protocol stacks)
In this section some protocol stacks that support this logical architecture are shown. As mentioned earlier,
the focus is on Ethernet as a common transport layer on top of any transmission technology, but other
solutions can also be considered. For that one example relying on SDH is also shown.
The first example is a pure IEEE 802.1ad based scenario. Service VLANs (S-VLAN) are used to carry
Customer VLANs (C-VLAN) across the Ethernet domain. One or more C-VLANs can be used (e.g. 1 per
eNB, different VLANs per different flows, etc.).
The Ethernet control plane is represented, for example, by protocols such as G.8031 (Ethernet Line
Protection), G.8032 (Ethernet Ring Protection) or standard Spanning Tree algorithms.
Figure 7: Example protocol stacks for Carrier Ethernet backhaul
GTP-U GTP-U
UDP UDP
IP IP
802.1Q 802.1Q (C) 802.1Q (C)802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
802.1Q (S) 802.1Q (S)802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S)
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
S1-AP S1-AP
SCTP SCTP
IP IP
802.1Q 802.1Q (C) 802.1Q (C)802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
802.1Q (S) 802.1Q (S)802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S)
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
802.1Q (C) 802.1Q (C) 802.1Q
802.1Q (C) 802.1Q (C) 802.1Q
eNB
CSG/Demarcation
node
Packet node -
First mile
Packet node -
Second mile
Packet node -
Aggregation MME
eNB
CSG/Demarcation
node
Packet node -
First mile
Packet node -
Second mile
Packet node -
Aggregation SGW
G.8031, G.8032,
LAG, STP

Access Aggregation


S1-U
S1-C
G.8031,
G.8032, LAG
G.8031, G.8032,
LAG, STP
G.8031, G.8032,
LAG, STP
G.8031,
G.8032, LAG
G.8031, G.8032,
LAG, STP
16
A second example is shown in the following figure. It is supposed that SDH is used in access and
aggregation (Ethernet over SDH model).
Figure 8: Example protocol stacks for Ethernet over SDH
GTP-U GTP-U
UDP UDP
IP IP
802.1Q 802.1Q (C) 802.1Q (C)802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q
802.1Q (S) 802.1Q (S)802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S)
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
GFP-F GFP-F GFP-F GFP-F GFP-F GFP-F
VCAT VCAT VCAT VCAT VCAT VCAT
SDH SDH SDH SDH SDH SDH
S1-AP S1-AP
SCTP SCTP
IP IP
802.1Q 802.1Q (C) 802.1Q (C)802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q
802.1Q (S) 802.1Q (S)802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S)
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
GFP-F GFP-F GFP-F GFP-F GFP-F GFP-F
VCAT VCAT VCAT VCAT VCAT VCAT
SDH SDH SDH SDH SDH SDH
802.1Q (C) 802.1Q
802.1Q (C) 802.1Q
S1-U
S1-C
G.8031 G.8031
G.8031
G.8032
G.8031 G.8031
G.8031
G.8032


Access Aggregation


17
5.3. Scenario 2 – Carrier Ethernet + L2/L3 VPN
This second scenario considers two different approaches for access and aggregation. The assumption is
that access is still based on a tree-like topology, while ring or mesh is introduced in aggregation.
While for access the same reasons for having Ethernet backhaul, as in scenario 1, are still valid, in
aggregation it is quite common to have MPLS and leverage its protection capability.
The next picture shows the high level architecture.
Core
First
mile Second Aggregation
mile
Service/
Networking
plane
Ethernet Ethernet Ethernet /DWDM
IP
IP/MPLS
Ethernet
S1
MEF EVC
Demarcation
node
Packet
node
Packet
node
Demarcation
node
L2/L3 VPN
VLAN-based Service Layer
Controller
Access Aggregation Core
eNB
Ethernet
any network
X2
Figure 9: Backhaul Scenario based on Carrier Ethernet with L2/L3VPNs
Specifically in aggregation, technologies such as DWDM can often be found together or as an alternative
to Ethernet. On top, an IP/MPLS based transport is used to carry LSPs and pseudowires, realizing a L2
or L3 VPN.
Since access is still based on an Ethernet, as per scenario 1, there is one node at the border between
access and aggregation that has the task of adapting the networking technologies and the functional
architectures.
As an alternative, MPLS-TP could also be employed in access instead of a VLAN transport. MPLS-TP
could also be used in aggregation for implementing a point-to-point L2 VPN (e.g. VPWS).
5.3.1. Applicability
Case 2 is one typical example for having a logical architectural split between access and aggregation.
The access can adopt any physical technology and rely on their protection methods (e.g. physical
redundancy). For the aggregation, the MPLS control plane handles all the necessary protection
mechanisms.
End-to-end OAM can be obtained at the pseudowire level with interworking with Ethernet OAM.
18
5.3.2. Considered implementations (protocol stacks)
Two examples are shown in this section.
In the first one a L2 VPN is considered.
Figure 10: Example protocol stacks for Carrier Ethernet with L2VPN backhaul
GTP-U GTP-U
UDP UDP
IP IP
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C) MPLS PW MPLS PW
802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S) MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
S1-AP S1-AP
SCTP SCTP
IP IP
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C) MPLS PW MPLS PW
802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S) MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation MME
802.1Q
VSI/Bridging VSI/Bridging
802.1Q 802.1Q 802.1Q
SGW
802.1Q
VSI/Bridging VSI/Bridging
802.1Q 802.1Q 802.1Q
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation
eNB CSG/Demarcation node
Control: BGP, IGP, RSVP-TE
Service: T-LDP
Control: BGP, IGP, RSVP-TE
Service: T-LDP


S1 -U
S1 -C
G.8031,
G.8032, LAG,
G.8031, G.8032,
LAG, STP
G.8031,
G.8032, LAG,
G.8031, G.8032,
LAG, STP
Access Aggregation
19
The second picture shows the same scenario where L3 VPN is used in aggregation.
UDP UDP
IP IP
802.1Q C) 802.1Q C) 802.1Q C) 802.1Q C) VRF VRF
802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S) MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
802.1Q C) 802.1Q C) 802.1Q C) 802.1Q C) VRF VRF
802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S) MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation SGW
S1-AP S1-AP
Routing (IP) Routing (IP)
802.1Q 802.1Q 802.1Q
GTP-U GTP-U
Routing (IP) Routing (IP)
802.1Q 802.1Q 802.1Q 802.1Q
802.1Q
Control: BGP, IGP,
RSVP-TE
Service: MP-BGP
Control: BGP, IGP,
RSVP-TE
Service: MP-BGP
S1-U
S1-C
G.8031,
G.8032, LAG,
G.8031, G.8032,
LAG, STP
G.8031,
G.8032, LAG,
G.8031, G.8032,
LAG, STP
Controller
Access Aggregation
eNB
S1-U
Figure 11: Example protocol stacks for Carrier Ethernet with L3VPN backhaul
5.4. Scenario 3 – MPLS access + L2/L3 VPN
This scenario mainly focuses on the transport capability of MPLS, which is also used in access.
Specifically, MPLS or MPLS-TP is considered to build point-to-point connections in the access domain as
a way to enter a VPN into the aggregation.
Core
First
mile Second Aggregation
mile
Service/
Networking
plane
IP
S1
MPLS/MPLS-TP
Controller
Access Aggregation Core
eNB
Ethernet Ethernet Ethernet / DWDM
IP/MPLS
Ethernet
MEF EVC
Demarcation
node
Packet
node
Packet
node
Demarcation
node
L2/L3 VPN
Ethernet
any network
X2
Figure 12: Backhaul Scenario based on MPLS with L2/L3VPN

 !
UDP UDP
IP IP
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C) VRF VRF
802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S) MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C) VRF VRF
802.1Q (S) 802.1Q (S) 802.1Q (S) 802.1Q (S) MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation SGW
S1-AP S1-AP
Routing (IP) Routing (IP)
802.1Q 802.1Q 802.1Q
GTP-U GTP-U
Routing (IP) Routing (IP)
802.1Q 802.1Q 802.1Q 802.1Q
802.1Q
Control: BGP, IGP,
RSVP-TE
Service: MP-BGP
Control: BGP, IGP,
RSVP-TE
Service: MP-BGP
S1-C
G.8031,
G.8032, LAG,
G.8031, G.8032,
LAG, STP
G.8031,
G.8032, LAG,
G.8031, G.8032,
LAG, STP

Access Aggregation


S1 -U
20
5.4.1. Applicability
Case 2 is probably the most suitable for adopting this logical architecture, especially if the need is to
leverage on the transport facilities of MPLS/MPLS-TP even in the access. In doing that, one or more
pseudowires carry the relevant traffic across the access network up to the first aggregation node where
traffic enters a L2 or L3 VPN.
End-to-end pseudowire OAM can be enabled through standard mechanisms such as VCCV ping,
traceroute or BFD.
5.4.2. Considered implementations (protocol stacks)
Again, two examples are reported here. The first one is based on a L2 VPN in the aggregation domain
(H-VPLS).
Figure 13: Example protocol stacks for MPLS with L2VPN backhaul
GTP-U GTP-U
UDP UDP
IP IP
802.1Q
MPLS PW 1 MPLS PW 1 MPLS PW 2 MPLS PW 2
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
802.1Q
MPLS PW 1 MPLS PW 1 MPLS PW 2 MPLS PW 2
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation SGW
802.1Q
VSI/Bridging VSI/Bridging
802.1Q 802.1Q 802.1Q
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation MME
S1-AP S1-AP
802.1Q
VSI/Bridging VSI/Bridging
802.1Q 802.1Q 802.1Q
Control: BGP, IGP
Service: (T -)LDP
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP -TE
Control: BGP, IGP,
RSVP-TE
Service: T -LDP
S1-U
Control: BGP, IGP
Service: (T -)LDP
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP -TE
Control: BGP, IGP,
RSVP-TE
Service: T - LDP
S1-C

Access Aggregation


21
The second examples only differentiate for having a L3 VPN in aggregation.
Figure 14: Example protocol stacks for Carrier Ethernet with L3VPN backhaul
UDP UDP
IP IP
802.1Q 802.1Q
MPLS PW 1 MPLS PW 1
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
802.1Q 802.1Q
MPLS PW 1 MPLS PW 1
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
802.1Q 802.1Q
VRF VRF
802.1Q
Packet node -
Aggregation MME
S1-AP S1-AP
Routing (IP) Routing (IP)
GTP-U GTP-U
Routing (IP) Routing (IP)
802.1Q 802.1Q
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation SGW
802.1Q
802.1Q 802.1Q
VRF VRF
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Control: BGP, IGP
Service: (T -)LDP
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP -TE
Control: BGP, IGP,
RSVP-TE
Service: MP -BGP

Access Aggregation


Control: BGP, IGP
Service: (T -)LDP
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP -TE
Control: BGP, IGP,
RSVP-TE
Service: MP-BGP
S1-U
S1-C
22
5.5. Scenario 4 – L2/L3 VPN in access + L2/L3 VPN in aggregation
Scenario 4 constitutes an interesting variation of scenario 3, where not only is the transport and protection
capability of MPLS exploited in access, but also the MPLS traffic segregation through VPN to have the
isolation of the two domains. In this case two distinct VPNs are provisioned, one in the access domain
and a second one into the aggregation.
Core
First
mile Second Aggregation
mile
Service/
Networking
plane
IP
S1
L2/L3 VPN
Controller
Access Aggregation Core
eNB
MPLS/MPLS-TP
Ethernet Ethernet Ethernet Ethernet / DWDM
L2/L3 VPN
Ethernet
any network
X2
Demarcation
node
Packet
node
Packet
node
Demarcation
node
IP/MPLS
Figure 15: Backhaul Scenario based on L2/L3VPNs in access and aggregation
5.5.1. Applicability
Several combinations are possible, using either IP/MPLS or MPLS-TP to obtain the different flavors of
VPN. The most likely see the presence of a L3 VPN in the aggregation, based on IP/MPLS, and a L2
VPN in access (even based on MPLS-TP), but for sake of completeness all the protocols stacks are
shown.
Case 3 and case 4 can be considered for implementing this scenario.
End-to-end OAM can combine LSP and pseudowire tools.
23
5.5.2. Considered implementations (protocol stacks)
A few possibilities are considered.
A first example shows a L2 VPN in the access domain combined with L3 VPN in aggregation.
Figure 16: Example protocol stacks for combined L2VPN and L3VPN backhaul
UDP UDP
IP Routing IP
802.1Q (C) VSI/Bridging 802.1Q (C)
MPLS PW MPLS PW
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP Routing IP
802.1Q (C) VSI/Bridging 802.1Q (C)
MPLS PW MPLS PW
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
S1-AP S1-AP
Routing (IP)
VSI/Bridging
VRF 1 VRF 1
802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile Packet node - Second mile
Packet node -
Aggregation MME
GTP-U GTP-U
Routing (IP)
VSI/Bridging
VRF 1 VRF 1
802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile Packet node - Second mile
Packet node -
Aggregation SGW
Control: BGP, IGP,
RSVP-T
Service: LDP, BGP
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE

Access Aggregation


Control: BGP, IGP,
RSVP-TE
Service: LDP, BGP
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, BFD,
RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, MP-BGP
Control: BGP, IGP, RSVP-TE
Service: LDP, MP-BGP
S1-U
S1-C
24
Another possibility is given by a double L2 VPN, as show in the next figure.
Figure 17: Example protocol stacks for two L2VPNs
UDP UDP
IP IP
Bridging Bridging 802.1Q (C)
MPLS PW MPLS PW MPLS PW MPLS PW
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
Bridging Bridging 802.1Q (C)
MPLS PW MPLS PW MPLS PW MPLS PW
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
S1-AP S1-AP
802.1Q (C)
Bridging (802.1xx) Bridging (802.1xx)
802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile Packet node - Second mile
Packet node -
Aggregation MME
GTP-U GTP-U
802.1Q (C)
Bridging (802.1xx) Bridging (802.1xx)
802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile Packet node - Second mile
Packet node -
Aggregation SGW
Control: BGP, IGP,
RSVP-T
Service: LDP, BGP
Control: BGP, IGP,
RSVP-TE
Service: LDP, BGP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP,
RSVP-TE
Service: LDP, BGP
Control: BGP,IGP,
RSVP-TE
Service: LDP, BGP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
S1-U
S1-C

Access Aggregation


25
At last, the case with two L3 VPNs is shown.
Figure 18: Example protocol stacks for two L3VPNs

UDP UDP
IP Routing Routing IP
VRF 1 VRF 1 VRF 2 VRF 2
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP Routing Routing IP
VRF 1 VRF 1 VRF 2 VRF 2
MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
S1-AP S1-AP
Routing (IP) Routing (IP)
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile Packet node - Second mile
Packet node -
Aggregation MME
GTP-U GTP-U
Routing (IP) Routing (IP)
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile Packet node - Second mile
Packet node -
Aggregation SGW
Control: BGP, IGP, RSVP-TE
Control: BGP, IGP, Service: LDP, MP-BGP
RSVP-TE
Service: LDP, MP-BGP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Control: BGP, IGP, Service: LDP, MP-BGP
RSVP-T
Service: LDP, MP-BGP
Control: BGP, IGP, RSVP-
TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVPTE
Service: LDP, RSVP-TE
S1-U
S1-C

Access Aggregation


26
5.6. Scenario 5 – End-to-end (Multi-segment) Pseudowire
Scenario 5 constitutes a variation of scenario 3 that combines the approach of having an end-to-end
service span (through different segments of the same pseudowire) yet maintaining possible different
transport and security domains (e.g. one in the access and another in the aggregation) by having an
access/aggregation demarcation node where all pseudowires are terminated or switched. MPLS(-TP)
LSPs or Ethernet VLANs can be used to steer and segregate the traffic, depending on Service Providers’
attitude.
Core
First
mile Second Aggregation
mile
Service/
Networking
plane
IP
MPLS / MPLS-TP
S1
MPLS/MPLS-TP/VLAN
MS-PW MS-PW
Controller
Access Aggregation Core
eNB
Ethernet Ethernet Ethernet Ethernet / DWDM Ethernet
any network
X2
Demarcation
node
Packet
node
Packet
node
Demarcation
node
Figure 19: Backhaul Scenario with multi-segment pseudowires
The picture considers only two segments for a pseudowire, for simplicity, but more are also possible.
5.6.1. Applicability
Case 2 and case 3 are well covered by this scenario, since different combinations of L2 and L3 transport
options are available, especially in access where Ethernet might be employed as an alternative to MPLS.
27
5.6.2. Considered implementations (protocol stacks)
The next picture highlights the stack associated with this scenario.
Figure 20: Example protocol stacks for multi-segment pseudowire backhaul
UDP UDP
IP IP
802.1Q MS-PW MS-PW MS-PW MS-PW MS-PW MS-PW
MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2 MPLS LSP 3 MPLS LSP 3
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
MS-PW MS-PW MS-PW MS-PW MS-PW MS-PW
MPLS LSP 1 MPLS LSP 1 MPLS LSP 2 MPLS LSP 2 MPLS LSP 3 MPLS LSP 3
eNB CSG/Demarcation node Packet node - First mile
Packet node - Second
mile
Packet node -
Aggregation SGW
GTP-U GTP-U
802.1Q (C) 802.1Q (C) 802.1Q
eNB CSG/Demarcation node Packet node - First mile
Packet node - Second
mile
Packet node -
Aggregation MME
S1-AP S1-AP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE,
L2TPv3
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP-TE,
L2TPv3
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP-TE,
L2TPv3
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE,
L2TPv3
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP-TE,
L2TPv3
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP-TE,
L2TPv3

Access Aggregation


S1-U
S1-C
802.1Q
802.3 802.3
802.1Q (C)
802.3 802.3 802.3 802.3 802.3 802.3 802.3
802.1Q (C)
802.3
802.1Q
28
5.7. Scenario 6 – Full L3
The name chosen for this scenario indicates a logical architecture based on MPLS/MPLS-TP as the
transport spanning from the eNB to the MME and S/P-GW controllers, on top of which one VPN is
provisioned to realize the service, be it a L2 or L3 VPN.
As a result, there is no logical distinction between access and aggregation, both belong to one domain
only.
Core
First
mile Second Aggregation
mile
Service/
Networking
plane
IP
MPLS / MPLS-TP
S1
Pseudo Wire
Controller
Access Aggregation Core
eNB
Ethernet Ethernet Ethernet Ethernet / DWDM Ethernet
Demarcation
node
Packet
node
Packet
node
Demarcation
node
any network
X2
Figure 21: Backhaul Scenario with full L3 network
5.7.1. Applicability
Case 4 is suited for this scenario, fully relying on MPLS or MPLS-TP for steering and protection.
This scenario will be very efficient for multi operator implementation.
29
5.7.2. Considered implementations (protocol stacks)
The first example shows an end-to-end L2 VPN.
Figure 22: Example protocol stacks for L2VPN based backhaul
GTP-U GTP-U
UDP UDP
IP IP
MPLS PW MPLS PW
MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
S1-AP S1-AP
SCTP SCTP
IP IP
MPLS PW MPLS PW
MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation MME
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation SGW
802.1Q 802.1Q (C) 802.1Q (C) 802.1Q
802.1Q 802.1Q (C) 802.1Q (C) 802.1Q
eNB CSG/Demarcation node
Control: BGP, IGP, RSVP -TE
Service: T - LDP
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP -TE
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP -TE
Control: BGP, IGP, RSVP-TE
Service: T - LDP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP -TE
Control: BGP, IGP, RSVP -TE
Service: LDP, RSVP -TE
S1-U
S1-C

Access Aggregation


30
This second case shows the routed version (L3) of this scenario.
UDP UDP
IP IP
VRF VRF
MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
VRF VRF
MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation SGW
GTP-U GTP-U
Routing Routing
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
eNB CSG/Demarcation node
Packet node - First
mile
Packet node - Second
mile
Packet node -
Aggregation MME
S1-AP S1-AP
Routing Routing
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
Control: BGP, IGP, RSVP-TE
Service: LDP, MP-BGP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, MP-BGP
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Control: BGP, IGP, RSVP-TE
Service: LDP, RSVP-TE
Controller
Access Aggregation
eNB
S1-U
S1-C
Figure 23: Example protocol stacks for L3VPN based backhaul
6. How to “pick” the right scenario
Even if this could be considered as the key question, the answer is not straightforward and does depend
on several factors.
Previous sections of this paper highlighted how the installed base, or even the incumbency of previous
technical choices, could be one of the drivers. In that case, operators handling the move from a TDM
network might prefer a stepwise approach and progressively introduce layer 2 or 3 solutions in the
aggregation before considering the transition in the access.
If this is the case scenarios 2 or 3 might represent the target network, even if a pure layer 2 network as
represented by scenario 1 has its own applicability.
Greenfield LTE operators or operators wishing to build a converged network also for enterprise and
residential services, might consider layer 3 oriented scenarios, e.g. from 4 to 6. In general, scenario 6 is
seen, at the current stage, as a kind of target architecture for the medium-long term.
In any case, each operator will base the choice considering not only technical aspects. Organization, skill
and attitude, service opportunity, available budget etc. will also influence the selection of the preferred
scenario, technology and products.
  Figure 23: Example protocol stacks for L3VPN based backhaul
UDP UDP
IP IP
VRF VRF
MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
SCTP SCTP
IP IP
VRF VRF
MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP MPLS LSP
802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3 802.3
node
node - First
node - Second
node -
Routing Routing
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
node
node - First
node - Second
node -
Routing Routing
802.1Q (C) 802.1Q (C) 802.1Q (C) 802.1Q (C)
IGP, RSVP -TE
MP -IGP, RSVP -RSVP -TE
IGP, RSVP-TE
RSVP -TE
IGP, RSVP -TE
RSVP -TE
IGP, RSVP - TE
MP -IGP, RSVP -TE
RSVP -TE
IGP, RSVP -TE
RSVP -TE
IGP, RSVP - RSVP-TE



31
7. Future Study Items
The following items have been proposed during the workgroup discussions as trends that might have an
impact on the backhaul network. The analysis done so far showed that this is not the case, but in future
that might change. For this reason they have been listed as areas where future studies might be directed
to.
7.1. LTE-A
LTE Advanced (LTE-A) is a preliminary mobile communication standard, formally submitted as a
candidate 4G systems to ITU-T in the fall 2009, and expected to be finalized in 2011. It is standardized by
3GPP as a major enhancement of the LTE standard.
LTE-A provides a toolbox of solutions improving radio performance, like enhanced MIMO, usage of Relay
BTS and Coordinated Multipoint Transmission (CoMP). As studies and standardization are still ongoing,
it is not possible to identify and quantify the implication of LTE-A on the backhaul network.
7.2. Security
There are two major differences compared to WCDMA with respect to transport security
 Air interface encryption of user plane traffic is terminated at the eNB, thus user plane traffic in the
LTE mobile backhaul network is not secured by Radio Network Layer protocols.
 Since the LTE network architecture is flat, adjacent base stations (X2) and core nodes (MME, SGW)
(S1) become directly IP-reachable from base station sites. If physical access to the site
cannot be prohibited, a hacker could connect his device to the network port and attack the
aforementioned network elements.
Transport security features are seen as mandatory if both the mobile backhaul network and the eNB site
cannot be regarded as secure. IPsec provides a comprehensive set of security features (data origin
authentication, encryption, integrity protection), solving both problems above. The 3GPP security
architecture is based on IPsec and Public Key Infrastructure (PKI).
IPsec is applied between Security Gateways (SEG). Each eNB site instantiates one SEG function. One or
more Security Gateways (SEG) should be located at the edge of the operator’s “Security Domain” (as per
3GPP Network Domain Security). Typically, the Security Domain includes the backbone network. The
SEG is effectively hiding the core nodes (MME, SAE-GW, Network Management System) from the mobile
backhaul network. Since the traffic itself is transparent to the SEG, various core network configurations
can be supported.
With the introduction of S1-Flex (one eNB connected to multiple EPC) also multiple SEG can be used,
which increases the number of security associations needed per eNB and also has some implication on
the architecture of the backhaul network itself.
The impact of security onto backhaul architectures will be anyhow addressed into a separate NGMN
paper.
32
7.3. Selective IP Offloading (SIPTO)
SIPTO is a concept being introduced in 3GPP Release 10 for reducing the “cost per bit” in flat networks. It
is based on a specific scenario within the operator’s network, allowing selective offloading of the traffic
away from the Evolved Packet Core network. One of its main goals is cost-optimized handling of the
internet traffic that is not intended for the operator’s core network (i.e., operator services).
Selective offloading can be based e.g. on the service type or specific QoS needs of the service.
Figure 24: Baseline approach for “SIPTO above RAN” scenario.
As shown in Figure 24 the Local GW is selected for the traffic to be offloaded. Hence for the backhaul
network (between eNB and S-GW) there is no implication caused by SIPTO.
7.4. Femto-Cells or none-3GPP Access
Femto-Cell base stations as well as none 3GPP-based radio products can share the same backhaul
network with the eNodeBs. In this case the corresponding security elements need to be included.
8. Conclusions and next steps
This paper has analyzed different topologies and architectures for an LTE-ready backhaul network and
aimed at supporting operators in their decision process towards a packet-based network capable of
supporting current and future requirements, as seen in the paragraph before.
The reason why several scenarios have been presented is that many options are possible, each with its
own advantages. In some cases operators will select one of the possible scenarios and will stay with it. In
other cases it is also possible to define a roadmap that includes more scenarios.
Before any architecture design, operators shall list their requirements to be considered, among others,
the installed base and how simple the migration to packet can be done, economical elements, the desire
to include factors not included in this paper such as security and high availability constraints or targeted
quality of experience.
33
9. References
1. NGMN, \'Next Generation Mobile Networks Beyond HSPA & EVDO – A white paper\', V3.0,
December 2006, available at www.ngmn.org.
2. “Next Generation Mobile Networks Optimised Backhaul Requirements”, NGMN Alliance, August
14th, 2008