Technical White Paper for

Seamless MPLS Networking

Contents
1 Preface.......................................................................2
1.1 Metro Bearer Technology in the TR101 Architecture..................... 2
1.2 Introduction of Seamless MPLS Networking Technologies............. 5

2 The Value of Seamless Networking Technology...........7

2.1 Unified and Simplified Bearer Technology..................................... 7
2.2 Flexible and Scalable Network Architecture.................................. 7
2.3 Quick Service Provision to Improve User Satisfaction..................... 9
2.4 Maturity and Reliability of IP/MPLS and OAM Technologies........... 10

3 Implementing the Seamless MPLS Networking

Technology.................................................................11
3.1 Challenge of the Seamless MPLS Networking Technology............. 11
3.2 Seamless MPLS Area-Based Networking....................................... 12
3.3 Seamless MPLS AS-Based Networking.......................................... 15

4 Key Technologies........................................................17
4.1 Inter-area LDP.............................................................................. 17
4.2 Service Driven LDP DoD................................................................ 18
4.3 MPLS Load-Balancing Label (MPLS Flow Label)............................. 18
4.4 Fast Convergence Technology...................................................... 20

4.4.2 LDP FRR Convergence Technology............................................. 20

4.4.3 BGP FRR Convergence Technology............................................ 21
4.5 OAM Technology......................................................................... 22

5 Applications................................................................24
5.1 Flexible Service Wholesale............................................................ 24
5.2 Inter-Metro VPLS Private Line....................................................... 24
5.3 ATM DSLAM Migration................................................................ 26

6 Conclusion..................................................................28
7 Appendix A References...............................................29
8 Appendix B Acronyms and Abbreviations....................30

Technical White Paper for

Seamless MPLS Networking
Keywords:

Seamless; MPLS; Scalability; Reliability; OAM; LDP DoD; PWE3

Abstract:

Seamless MPLS networking refers to the formation of a unified IP/MPLS

control plane for all the IP devices managed by operators, including
access (fixed/mobile), convergence, and backbone devices. The
Seamless MPLS networking architecture greatly reduces cooperation
between different network layers when services are deployed, allows
operators to quickly provide services, and lowers deployment costs.
Reliability/OAM detection can be deployed easily through the end
to end IP/MPLS control plane and unified networking technology.
This document describes the background, challenges, solution, and
application scenarios of the Seamless MPLS.

1 Preface
With the wide application of Multiprotocol Label Switching (MPLS)
technology, operators establish Metropolis Area Networks (MANs) and
backbone networks for the bearing of integrated services by using
the MPLS-based L3VPN/L2VPN solution. Generally, services and users
are identified through 802.1Q or 802.1ad between access devices,
MAN, and backbone networks, requiring the static configuration of
interfaces between each layer. In seamless networking, the end-toend IP/MPLS networking of all data devices managed by operators is
realized, including access (fixed/mobile), convergence, and backbone
devices. This is achieved by eliminating the existing 802.1Q/802.1ad
interface between devices on each layer in the existing network. Thus,
each layer can implement dynamic interaction, which is called end-toend big networking. This chapter describes the problems with existing
networking architecture, and the benefits of Seamless MPLS architecture.

1.1 Metro Bearer Technology in the TR101 Architecture

The TR101 in the DSL forum draft specifies the DSL aggregation model
in Ethernet mode.

NSP/BB
Network
Gateway

NSP1
NSP2
ASP1

A10-NSP
L2TP

A10-NSP

L2TS

IP - QoS

A10-NSP

IP

BB
Network
Gateway

IP - QoS

A10-NSP Regional Broadband

Ethemet
Aggregayion

Access
Node
(DSLAM)

Access
MDF Loop

V
Access Network

Network

User1
ND

CPE
T

User2

U Customer Prem.Net

Aggregation Network
Figure 1 TR101 Ethernet DSL aggregation model

The model defines the V interface between the access node and
Ethernet convergence network/node to identify different DSL ports
for services and user access through the 2-layer TAG of 802.1ad
(QinQ).
The mainstream networking technologies of the metro convergence
network include Ethernet enhanced technology (QinQ, PBB), MPLS
bearer, and L3 Hybrid. The Ethernet enhanced modes, such as QinQ
and PBB, effectively improve network reliability and service flexibility.
The MPLS bearer mode is one of the mainstream bearer technologies
of the Ethernet convergence network, as it facilitates VLAN scalability
and reliability. In L3 Hybrid mode, services are classified into edge
processing services and transparently transmitted services, according
to service features. For the edge processing services, the IP edge is
located in the edge convergence node. Transparently transmitted
services are sent to the specified POP point through the MPLS pipe.
The Hybrid mode can adapt to service development requirements in
the future.

PPPOE VLAN

VPLS
VPLS
VPLS
VPLS
VPLS

AGG

UPE

PPPOE VLAN
VOD VLAN
IPTV VLAN
L2VPN VLAN
L3VPN VLAN

VPLS

IPTV VLAN
L2VPN VLAN
L3VPN VLAN

AGG
A
GG

VOD VLAN

UPE

DSLAM/OLT
DSLAM/OLT

L3 Hybrid
Mode

MPLS
Mode

In these metro convergence network technologies, the V interface

specified in TR101 architecture is used. The following figure shows
the bearer modes of services in the network position in MPLS mode
and L3 hybrid mode.

PPPOE VLAN
VOD VLAN
IPTV VLAN
L2VPN VLAN
L3VPN VLAN

BRAS

PPPOE VLAN
L3VPN

BRAS

IP Multicast
L2VPN
L3VPN

SR
PE

SR

IA,Native IP
L3VPN
IP Multicast

IXC
CS
VHO

L2VPN
L3VPN

IA, Native IP

IXC
CS
VHO

PF

Figure 2 Service bearer model of the TR101 architecture

With the continuous development of user access requirements and

services, network operators must meet these new requirements.
TR101 architecture cannot meet these new requirements.
1. For FTTX access, the wholesale service mode of the ULL cannot be
maintained. POP point locations vary with the scale of alternative
operators. Therefore, incumbent operators must provide a more
flexible connection capability at the access node (AN). In this case,
the AN must identify alternative operators, services, and users.
For TR101 architecture, the complicated VLAN planning is required
on the V interface due to the limitation of the VLAN space. In
a typical metro MPLS networking scenario, the VLAN must be
configured in more nodes (AN, UPE, AGG, and PE) of incumbent
operators if the POP point location of alternative operators is not in
this local MAN.
2. Connecting the inter-metro enterprise private line is required due
to the quick development of the enterprise Ethernet private line.
Multi-point static configuration in the TR101 architecture is
required. The inter-metro Ethernet private line can be provided
through cooperation between the management entities on each
layer.
3. According to Fixed-Mobile Convergence (FMC) requirements, the
AN must be accessed to the mobile bearer services to connect the
base station and mobile gateway, including ATM, TDM, and the
Ethernet.
4. The function of the IP bearer network is more important due to
service access density and the migration of telecom services. Users
require greater access reliability.
In TR101 architecture, the networking technologies on each
layer are not unified and different reliability solutions are used.
Cooperation and interaction are complicated without a mature
standard. Device manufacturers provide solutions of different
reliability and the interoperability problem cannot be solved.

1.2 Introduction of Seamless MPLS Networking

Technologies
With Seamless MPLS architecture, the entire network uses unified IP/
MPLS networking technology, with an end-to-end control plane. As a
result, the V interface between the AN and Ethernet convergence is
eliminated. The above problems are solved. Through Seamless MPLS
networking technology, operators can flexibly handle connection
requirements so that services are provided quickly. Thus, new services
are deployed quickly and service deployment costs are reduced, which
strengthens the competitiveness of operators.
In addition to the V interface between the access and convergence
network, the VLAN may be configured for service interconnection
between the MAN and backbone network. Seamless MPLS networking
can meet this requirement and enable dynamic establishment through
the end-to-end MPLS pipes for inter-metro services.
In the Seamless MPLS networking model, all services can be
transferred to the specified service processing points through the MPLS
pipe at service access points. The entire bearer solution is simple and
consistent. With the dynamic end-to-end MPLS pipe establishment
capability, services can be transferred to any service processing points
or peer service access points through the MPLS pipe/PWE3 pseudo
wire, according to requirements.

Access

Metro Aggregation

Metro Aggregation

Access

MPLS Domain

OLT/DSLAM
BNG

SRG
CSG

IP Backbone

LER

LER

OLT/DSLAM
SRG

BNG

LSR
UPE

AGG

AGG

UP

CSG

MPLS
IGP FC / MPLS OAM / MPLS FRR
Figure 3 Seamless networking model

Seamless MPLS networking technology is not new and is widely

applied to backbone networks, MANs, and the mobile backhaul.
Application experiences indicate that Seamless MPLS networking is a
mature and reliable bearer technology with excellent scalability. The
Seamless MPLS can eliminate the gap between network layers to
implement end-to-end MPLS networking.

2 The Value of Seamless

Networking Technology
2.1 Unified and Simplified Bearer Technology
With seamless networking technology, the access (fixed/mobile),
convergence, and backbone devices connect through the unified IP/
MPLS control layer.

2.2 Flexible and Scalable Network Architecture

Future-oriented networking technology must provide a more flexible
and scalable network architecture. Currently, the development of
some services indicates that more flexibility and better scalability are
required.
1. Service wholesale in FTTX access
In the case of the copper access, the LLU is the main wholesale
mode. In the case of the FTTX access, the LLU is not applicable. The
bit stream access will become the mainstream wholesale mode.

Network Open Point

Copper Access Mode

Local
Exchange

Customer
Copper Access

1) Copper ULL

Metro POP

2) PVC/VLAN Bit stream

Street/
Building

3) L2TP Wholesale

Network Open Point

FTTx Access Mode

Customer

IP Core

Local
Exchange

Metro POP

IP Core

Figure 4 Service wholesale in the copper and FTTx access scenarios

The POP device can be deployed in any network layer according to

the network scale of the competitor operator. For BSA wholesale, the
network must be flexible and simple so that wholesale user services
can be handed over to a competitor operator at any point. Through
end-to-end MPLS pipes between the ANs and service transfer points,
seamless architecture meets flexible connection requirements. In
TR101 architecture, the AN, convergence devices, and service transfer
points are configured. When the service transfer points are in the
inter-metro area, the distributed nodes and backbone PE devices in the
MAN must be configured.

TR101

Sometimes, the competitor operator needs to identify different users

and services through the 2-layer VLAN TAG. To meet this requirement,
the incumbent operator must identify the competitor operator and
its users and services on the AN. If TR101 architecture is used, the
distribution policy of the S-TAG and C-TAG must be planned carefully.
In the seamless architecture, the PW identifies different competitor
operators. It is easier to identify users and services.

Identify the
wholesale
operators,services,
and users

End user

AN

Identify the
pipe for
entering the
metro by VLAN

QinQ

UPE

Leave the metro

pipe, use the VLAN
to identify the
wholesale operator
again

PE-AGG

QinQ

Enter the
backbone
pipe by
VLANs

PE

Seamless

Identify services and

users. Enter the
PWE3 pipe
according to
wholesale operators.

Leave the
backbone pipe

PE

QinQ

Third Party
router

Leave the
pipe

QinQ
End user

PE

AN
MPLS Tunnel/PWE3

Third Party
router

Figure 5 Service wholesale deployment comparison between the seamless and TR101

2. Deployment of inter-MAN enterprise Ethernet private services

Enterprise Ethernet services are not limited just in the MAN. The
connection of Ethernet private lines may be required between
MANs. In TR101 architecture, multiple nodes such as access
nodes, metro UPE, metro PE-AGG, and the backbone PE must be
configured to provide Ethernet private line services between the
8

inter-metro DSLAMs. In addition, the VLAN planning is required.

In seamless networking, only the access nodes on both sides are
configured in the dynamic PWE3 mode to provide Ethernet private
line services.

TR101

Service configuration point

Metro 1
CPE

AN

UPE

PE-AGG

core
PE

Metro 2
PE

PE-AGG

UPE

AN CPE

UPE

AN CPE

Seamless

Service configuration point

Seamless MPLS
CPE

AN

UPE

PE-AGG

PE

PE

PE-AGG

Figure 6 Comparison of deployment of inter-metro private line services between the Seamless and TR101

2.3 Quick Service Provision to Improve User

Satisfaction
To provide wholesale and enterprise private line services under TR101
architecture the access, metro, and backbone devices of different
layers must cooperate. For service deployment, cross-departmental
coordination is required between the management entities of different
layers. As a result, it takes a long time to provide services.
With seamless networking, operators need to only configure user
access points. Services can be provided quickly, user satisfaction
increases, and service deployment costs are reduced. Thus, operators
competitiveness is strengthened.

2.4 Maturity and Reliability of IP/MPLS and OAM

Technologies
Based on IP/MPLS technology, seamless networking easily implements
end-to-end protection through the reliable and mature IP/MPLS
technology and OAM. The interoperability between the devices of
different manufacturers is excellent.
In TR101 architecture, protection between each layer requires both
IP/MPLS and Ethernet reliability to cooperate with OAM technology.
Deployment is complicated and interoperability may fail.

10

3 Implementing the Seamless

MPLS Networking Technology
3.1 Challenge of the Seamless MPLS Networking
Technology
Through IP/MPLS technology, the seamless MPLS connects the access
layer, convergence layer, and backbone layer, and provides flexible
and scalable networking architecture for operators. It is improper to
directly inherit all technologies from the old IP network.
After the devices of each layer are seamlessly connected, the scale of
the IP/MPLS domain improves by orders of magnitude compared with
the original networks. For example, in a network with 20,000,000
users, if each DSLAM connects 100 users in FTTC access mode, the
number of nodes in the entire network is over 200,000. If each OLT
connects 1000 users in FTTB/FTTH access mode, the number of nodes
in the entire network is 20,000. In the original networking mode,
the order of magnitude of the number of nodes in the backbone
and metro route domains is in the 1,000s. Hence, the scale of the
route domain in Seamless MPLS networking increases by an order of
magnitude of one or two. In a large-scale network, engineers have
to consider how to construct the route and MPLS tunnel, and how to
guarantee the availability of the networks.
In addition, a large number of access devices, such as DSLAMs
and OLTs, are available in the network, taking up a high ratio of
network investment. Hence, the introduction of the IP/MPLS should
not obviously affect the cost of access devices. In Seamless MPLS
networking, the complexity of the access device control plane and
performance specifications of the forwarding layer must be reduced.
The following table lists the typical specification of the IP/MPLS
capacity for access devices by a European telecom operator.
Parameter

Routing IP FIB LDP

Protocol
control
layer

Specifications Static
route

200

LDP
BGP
forwarding
table
200

Not
supported
11

3.2 Seamless MPLS Area-Based Networking

In the area-based networking mode, all devices belong to the same
autonomous system (AS). The IGP (OSPF/ISIS) is used to exchange
route information between nodes. The devices of each metro are
divided into different IGP areas. The IP backbone devices constitute
backbone areas or a level-2 area to ensure that the number of nodes
in each area is appropriate. The access node may adopt static routing,
without supporting the dynamic IGP protocol.
As shown in Figure 7, the entire route domain of the Seamless MPLS
is divided into three layers: the backbone node, metro convergence
node, and AN node. Serving as the ABR, the PE in the backbone
network converges area routes and advertises them to other areas,
which reduces the route capacity requirements of each device. Besides
reducing the routing table capacity of each device, route area-based
deployment is helpful for isolating inter-area faults and enabling fast
route convergence.

UPE

PE-AGG

Core-PE

Core-PE PE-AGG UPE

Core-P

AN

Static
route

ISIS Level -1 /
OSPF Area n

ABR

ISIS Level -2 /
OSPF Area 0

AN

Figure 7 Seamless MPLS route area-based deployment

PW labels are distributed in end-to-end T-LDP mode. Tunnel labels

are distributed in hierarchical mode. See Figure 8. In the dynamic
IGP range, the LDP DU label release mod is used. The LDP DoD label
release mode is used between the UPE and AN. The AN must be costeffective; therefore, DoD mode allows the AN to request the needed
labels on demand. As a result, the specification requirement of the
MPLS forwarding table is reduced. The LDP DU labels are distributed
to the edge of the area according to a certain policy. In the backbone
area (or level-2), the route label is distributed to the common area
(level-1). The route label of common areas is not distributed to the
12

backbone area. In this way, the UPE can establish the LSP tunnel to the
edge direction in any area.
The inter-area labels are distributed in Labeled BGP (RFC 3107) mode.
The iBGP runs through the UPE. Many UPEs are available; therefore,
the 2-level reflector structure is used. The Core-PE functions as the
UPE reflector of this area, and the core-RR functions as the Core-PE
reflector. The Core-PE is not a simple reflector. Upon receipt of the BGP
Label from the UPE, the Core-PE changes the Next-hop of this label
route information to the Core-PE, and re-allocates the labels. When
the Core-RR receives the label route information, it reflects it to the
UPE, without changing any information.

IBGP Neighbor
UPE

PE-AGG

BGP Lable for AN1

AN

Core-PE

BGP NH unchanged

LDP DU

Core-PE PE-AGG UPE

BGP NH unchanged

BGP NH Self

ABR
LDP DoD

Core-RR

BGP NH Self

Core-P

Lable for AN2

AN

ABR

LDP DU

LDP DU

Figure 8 Seamless MPLS label release area-based deployment

Figure 9 shows a label release and forwarding instance. The PW label

is directly allocated through the T-LDP session between DSLAM-A
and DSLAM-B. The external LSP tunnel from DSLAM-A to DSLAM-B is
actually divided into four segments:
Segment 1: LDP DoD label from the DSLAM-A to the UPE-A, which
is the DSLAM-B label requested by the DSLAM-A from the UPE-A on
demand. According to the self BGP label table, the UPE-A searches
the DSLAM-B to allocate to the DSLAM-A LDP label and establish the
matching relation between them (many-to-one).
Segment 2: Two-layer tunnel from the UPE-A to ABR-B. The external
layer is the tunnel established in the LDP DU mode from the UPE-A to
the ABR-B. The internal layer is the tunnel distributed to the DSLAM-B
13

through the Labeled BGP for the UPE-A by the ABR-B. The labels of the
external tunnel are changed hop by hop during forwarding. The labels
of the internal tunnel are invisible between the UPE-A and ABR-B, and
remain unchanged.
Segment 3: Tunnel from the ABR-B to the UPE-B. This segment is
similar to segment 2, which is a two-layer tunnel.
Segment 4: Tunnel from the UPE-B to the DSLAM-B. This tunnel is
obtained through the UPE-B request to the DSLAM-B in LDP DoD
mode according to the static route.

DSLAM-A
LDP
Labled BGP

Metro A

RR

UPE-A PE-AGG-A ABR-A

LSR

PE-AGG-B UPE-B DSLAM-B

DSLAM - B
LBL:8000

Require
DSLAM - B
ABR - B
LBL:7000

ABR - B
LBL:6000

ABR - B
LBL:5000

8000
50

7000
200
50

6000
200
50

PW -ID:100, LBL :50

3(NULL )
5000
200
200
50
50

LDP DU LSP
DoD Label

ABR - B
LBL:4000

DSLAM - B,LBL:200

DSLAM - B,LBL:200

PW Label
50

ABR-B

Require
DSLAM - B

T-LDP
Forwarding

Metro B

BGP Label 200

UPE - B
LBL:3000

UPE - B
LBL :2000

DSLAM - B
LBL:1000

DSLAM - B,LBL:100

3000
100
50

3(NULL)
100
50

LDP
DU
BGP Label 100

3(NULL)
50

DoD Label

Figure 9 Seamless MPLS control and forwarding instance (uni-directional)

In the above route and label release mode, the number of routes
and labels of each node are reduced. The entire networking solution
features excellent scalability. The following table lists the node route
and labels of the above solutions.

14

Node
IGP
route

Description

AN

UPE

AGG

Core-PE

Core-P

Default
static
route

Specific routes
in this area.
Convergence
route of other
areas

Specific
routes in this
area.
Convergence
route of
other areas

Specific routes
in this area and
backbone area.
Convergence route
of other areas

Specific routes in
the backbone area.
Convergence route
of other areas

2000

2000

3000

1000

All nodes in
this area.
All nodes in
the backbone
area.

All nodes in this

area.
All nodes in the
backbone area.

All nodes in the

backbone area.

Typical value 2
LDP
label

BGP
label

Description

Request All nodes in

on
this area.
demand All nodes in
the backbone
area.

Typical value 200

3000

3000

3000

1000

Description

None

All nodes

None

Control layer: all

nodes
Forwarding layer;
all nodes in this
area.

None

Typical value None

200,000

None

2000 (forwarding
layer)

None

3.3 Seamless MPLS AS-Based Networking

In the AS-based networking mode, each metro and backbone are
in different AS domains. The metro can use the private AS number.
The IGP protocol is independently deployed on each AS domain.
EBGP switching route information is used between the metro and
backbone area. When the EBGP advertises route information, routes
are converged. As a result, the number of routes decreases. Route
deployment of the access nodes is the same as area-based networking;
that is, static route mode.
Label distribution is similar to that in area-based networking. The LDP
DU mode is used in the AS-domain. In the inter-AS domain, labeled
BGP is used to release label routes. The DoD is used between the AN
and UPE. In ASBR position, the BGP labels perform bidirectional NextHop Self operations. The labels are re-allocated locally. Therefore, the
requirements for BGP label forwarding table capacity at the metro
egress and Core-PE position is high.

15

Figure 10 shows the distribution of routes and labels in AS-based

networking.

iBGP Neighbor

EBGP Neighbor
Core-P

AN

MetroA Mero egress Core-PE

Static UPE
AS 65001
route
ISIS level-2

LDP DoD

LDP DU
BGP NH Self

Core
AS 3000
ISIS level-2

LDP DU

Core-PE

UPE
Metro B
AS 65002

AN

LDP DU

BGP NH Self

Figure 10 Seamless MPLS AS-based route/label release

16

4 Key Technologies
4.1 Inter-area LDP
According to the LDP specification RFC5036, the IP address
information in the FEC should be checked when the LDP label route
information is received. When the IP address matches the route
information in the local IP route table, the label information is valid.
Actually, the above solution does not comply with this specification.
In the UPE location, the LDP LSP to all backbone nodes must exist.
When the route is advertised, the ABR aggregates the routes of the
backbone area. In the UPE, the specific route to the backbone node is
unavailable. The address in the LDP FEC fails to locate the accurately
matched route information. At the access nodes, only the default
route is configured. The label information of any node must be
requested on demand.
To meet Seamless MPLS networking requirements, the LDP Extension
for Inter-Area Label Switched Paths (RFC 5283) extends the restriction
in the LDP specification. The precision matching principle is changed
to the maximum length matching principle. If the address prefix in the
FEC has the maximum matching item in the IP route table and the next
hop of both is consistent, the label information is valid. The extension
supports the hierarchical deployment of the Seamless MPLS inter-area
route and labels.

17

4.2 Service Driven LDP DoD

On the AN nodes, the number of LDP label tables must be restricted.
If the LDP DU mode is used, the filtering policy must be configured on
the AN nodes. In this mode, a large number of policy tables must be
maintained. When the policy is changed, a mechanism is unavailable
to notify the upstream UPE to re-transmit the LDP label information.
The Downstream on Demand (DoD) LDP distribution mode meets the
Seamless MPLS AN node requirement. The labels can be requested
dynamically according to service requirements. When label forwarding
tables are reduced, flexible service change requirements are met.
Label distribution in DoD mode requires the AN node to be configured
with the request policy. When services are changed, the DoD request
policy must be changed, along with the service configuration.
Configuration and maintenance workloads increase. The service
driven LDP DoD can meet the AN nodes future flexible connection
requirements to automatically configure DoD request policies and
avoid the repeated configuration workload.

4.3 MPLS Load-Balancing Label (MPLS Flow Label)

In TR101 networking architecture, packets between the access
node and UPE are encapsulated in VLAN/QinQ mode. The load can
be balanced according to user MAC addresses and IP addresses.
After migration to the Seamless MPLS networking architecture, the
UPE functions as the P node. Current load-balancing technology is
implemented based on PW granularity. The number of PWs on the
access node is limited and PW traffic is not balanced. Hence, loadbalancing based only on PW granularity causes seriously unbalanced
loads. As a result, bandwidth scalability between the access node and
UPE is affected. The following table lists the typical services of the
AN. The bandwidth distribution of each service is unbalanced. The
corresponding PW bandwidth of the HIS service takes up more than
70% of the total bandwidth. PW granularity cannot implement loadbalancing.

18

Services Bandwidth

Description

PW1

800M

HIS service. Connects to the BRAS.

PW2

200M

IPTV live service. Connects to the SR.

PW3

100M

Wholesale service. Connects to the SP BRAS

PW4

10M

Enterprise private line 1. Connects to other

DSLAMs in the domain

PW5

2M

Enterprise private line 2. Used for the interdomain connection

When the single stream is carried in the PW, the packet load in the PW
is balanced (ECMP or Trunk) to multiple links. As a result, packets may
be disordered and user services affected. When the single PW carries
large traffic (for example, PW corresponding to the HIS, containing
the online traffic of a large number of users), all packets of each traffic
are carried over the same link, without affecting the services of other
users.
In the current load balancing mechanism, it is difficult for node P
to perform load-balancing by traffic on the PW. Hence, the MPLS
Flow Label is added to the ingress PE of the PW. The PE node can
identify the flow, with the understanding of the services carried on
the PW. Intermediate nodes need to only balance nodes by label, so
implementation is simplified. After the flow label is added, the packet
encapsulation of the PW is as follows:
LDP Label
BGP Label
PW Label
MPLS Flow Label
Control Word (optional)
Payload
Node P performs load-balancing according to the L4 label stack
information. The specific flow L4 label information in the PW is the
same, which ensures that all packets in the flow are not dissembled
to different egresses. In the PW, various flows can be balanced to
different links because the flow labels are different. In this way, the
PW is balanced according to the load of flows.
When the carried MPLS Flow Label needs to be established on the
PW, both ends need to notify each other about various issues, such as
whether to support the Flow Label, and whether the packets carry the
19

Flow Label. In this way, basic forwarding is not affected when either
party does not support the corresponding function.

4.4 Fast Convergence Technology

The Seamless MPLS networking adopts unified IP/MPLS technology.
The mature and reliable IP/MPLS technology enables end-to-end fast
protection switching when different vendors devices are networked.
The service protection process in the Seamless MPLS networking is
described below.
The network and bearer of the Seamless MPLS are hierarchical.
Accordingly, the protection technology is hierarchical. For example, in
the area-based networking solution, the LSP tunnel between the ANs
is divided into four segments. The roles of the tunnels in the node vary
with links. Hence, the convergence modes are different in the case of
failure.

traffic
LDP DU LSP
BGP Lable
Lable

LDP
LDP DoD
DoD
AN

UPE1

PE-AGG

ABR

LDP DU LSP
BGP Lable
Lable
P

Remote
ABR1

Remote PE-AGG

LDP DoD
Remote
UPE1

Remote
AN

Figure 11 Roles of the nodes in the Seamless MPLS area-based solution

As the basis for the reliability of Seamless MPLS networking, the fast
convergence technology of the IGP and BGP guarantee quick path
recovery in the control layer. In the forwarding layer, FRR technology
can accelerate service convergence.

4.4.2 LDP FRR Convergence Technology

In the LDP layer, LDP FRR technology implements the fast convergence
of the LDP LSP. In the LDP FRR technology, the device uses the optimal
route of the LDP as the forwarding entry, simultaneously uses the
second best route of the LDP as the backup path, and stores it in the
forwarding table. When the optimal next hop is faulty, the backup
path/label is used directly for forwarding.

20

The BFD detection technology quickly detects the connection of the

optimal next hop to implement the convergence rate of 50 ms.
The usage of the LDP FRR convergence technology is limited. For
example, in the ring networks, the second best next hop sends packets
to this node. As a result, the forwarding ring is formed.
For dual-homing networking of the AN, the ring does not exist. In this
case, LDP FRR technology can be used.
The networking of the nodes above the UPE is complicated. The ring
may exist. In this case, the Loop Free Alternates (IGP LFA) defined in
the RFC5286 is used to check whether the ring exists. The LFA is used
to check whether the second best path has a ring according to IGP
link status information. If the LFA detects a ring, the node does not
use LDP FRR convergence technology. In most cases, the LDP FRR can
be applied.
In comparison to the FRR technology of the RSVP TE, LDP FRR
protection is single point, not end-to-end. The LDP FRR can be
deployed if there are a large number of tunnels in Seamless MPLS
networking. TRSVP TE FRR protection technology may be deployed for
certain services or on certain network layers.

4.4.3 BGP FRR Convergence Technology

The LDP FRR rapidly switches paths if a link fails. When the BGP node
(including the UPE/ABR) is faulty, it depends on the convergence of the
label BGP. For common BGP convergence technology, convergence
is performed in the control layer and then the forwarding entry is
delivered. The entire convergence time may reach the second level.
The BGP next hop separation technology can increase the convergence
speed of the control layer. The carrier-class reliability requirement is
not met.
The BGP FRR adopts the direct switching mode in the forwarding layer.
The LDP Label/BGP Label of the second best BGP neighbor is directly
stored in the forwarding list as the backup. When the fast detection
mechanism (such as the BFD) detects that the best BGP neighbor
is faulty, the system directly switches to the backed up entry for a
quicker convergence performance.

21

The following table summarizes the convergence technology and

performance of the unidirectional flow if a fault occurs.

Fault
Location

AN-UPE
UPE-Remote Remote ABR
Link or UPE Inter-ABR Link Node
Node
or Node

Fault
detection

AN-UPE
BFD

Single-hop
UPE-remote
Single-hop
BFD detection ABR multi-hop BFD detection
between
BFD detection between points
points

Remote ABR
- Remote UPE
multi-hop
BFD detection

Remote UPE
- Remote
AN BFD
detection

Switching
Operation

LDP FRR

LDP FRR
IGP FC

BGP FRR

LDP FRR
IGP FC

BGP FRR

BGP route
cancellation

50 ms

200 ms

50 ms

200 ms

Subsecond
(only the
service of
this AN is
affected)

Convergence 50 ms
Performance

Remote ABR Remote UPE

Remote inter-UPE Node
Link or Node

4.5 OAM Technology

In the unified IP/MPLS control layer, the entire network is maintained and managed
through consistent OAM technology, which greatly reduces the inter-department
coordination time and the definition of inter-department responsibility. The fault can
be quickly and proactively detected, located, and handled, which improves QoE.
The entire network is deployed with hierarchical OAM. On the link layer, the linklevel fault can be detected and located through detection technology on the link.
On the tunnel layer, tunnel connectivity is detected online in real time through
the BFD for LSP, and MPLS OAM ( Y.1711). LSP ping/traceroute can locate failures
and faults in service provision. MPLS-TP-OAM performance monitoring technology
monitors the performance indexes of tunnel bearer services in real time. On the PW
layer, the VCCV monitors the connectivity of the PW in real time. The MPLS-TP OAM
for PW and Y.1731 for PW/VPLS monitors the specific service performance indexes
in real time. For the OAM packets between user devices, unified MPLS technology
easily implements transparent transmission.

22

Remote UPE
- Remote AN
Link

Seamless MPLS

CPE

AN

UPE

PE-AGG

Core

Core

PE-AGG

UPE

Link-level OAM

AN

CPE

MIP

Tunnel-level OAM,connectivity,BFD/MPLS OAM,performance MPLS-TP OAM

PW OAM,connectivity:VCCV,performance:MPLS-TP OAM for PW,Y.1731for PW/VPLS
End-to-end transparently transmitting the OAM packets

Figure 12 Unified OAM deployment

23

5 Applications
5.1 Flexible Service Wholesale
The Seamless MPLS architecture meets flexible service wholesale
requirements, as shown in Figure 13. Service transfer points vary
with the competitor operator, and the AN nodes need to transport
user flows to the corresponding transfer points according to the
competitor operators. In Seamless MPLS networking architecture, the
AN node can directly establish the PW connection with the service
transfer point, and be perceived by intermediate devices. Deployment
is simple. According to the requirements of competitor operators, user
location and the service label can be identified on the AN. The VLAN
flags of different competitor operators can be overlapped without
bottlenecking the number of competitor operators or users. End-toend service protection measures can be conveniently implemented.

Competitor operator1

Competitor operator3
User 3

User 1

PW3
PW1

Core -PE

Users
DSLAM/OLT

PW2
AGG

User 2

Seamless
MPLS
Competitor operator2

Figure 13 Deployment of flexible service wholesale

5.2 Inter-Metro VPLS Private Line

In the Seamless MPLS networking architecture, the VPLS service can
be conveniently deployed across the entire network without the
restriction of the VLAN.
A large number of nodes are available in the entire network. The VPLS
must be hierarchically deployed. In this way, this avoids the problems
24

related to full neighbor connections and the forwarding efficiency of

the broadcast/unknown packets.
See Figure 14. The VPLS service is deployed in the 3-layer H-VPLS
architecture. The H-VPLS core layer is constructed between the ABRs.
The UPE is the convergence layer, and is accessed to the ABR of this
area through the PW. For the CPE accessed in the DSLAM, the DSLAM
is accessed to the UPE through the PW. UPE location can be directly
accessed to the CPE in Ethernet/VLAN mode. In 3-layer architecture,
the number of PWs converged in each layer is limited; for example,
the UPE is accessed to 100 DSLAMs, and the ABR is accessed to 20
UPEs. If the DSLAM is directly connected to the ABR through the PW,
the pressure of the PW on the ABR is high. The topology of the PW is
close to the actual physical topology to reduce PW overlapping on the
single link, which improves the forwarding efficiency of the broadcast/
unknown unicast packets.
To reduce the workload of adding the service access points in the VPLS
instance and ease VPLS deployment, the BGP Auto-Discovery H-VPLS
can be deployed in the core and convergence layer nodes (ABR and
UPE) to automatically discover the new service access points. The new
service access points are processed between the DSLAM and UPE in
static configuration mode. When the DSLAM does not support the
BGP, you can configure only the DSLAM and UPE to provide services in
the scenario of adding the single VPLS service access point.
In the H-VPLS, the ABR is the core node. If many VPLS instances are
accessed, the pressure on the device MAC address table specification
is high. PBB+H-VPLS technology can handle the problem. In an actual
deployment scenario, the PBB can be encapsulated on the UPE. For
the packets reported through the PW by the DSLAM or the packets
reported through the Ethernet by the CPE, the user MAC addresses
(C-MAC) are filtered. The MAC address (B-MAC) of the destination UPE
is visible only on the ABR. The number of visible MAC addresses of the
ABR is greatly reduced. The scalability of VPLS service deployment is
improved. For PBB and VPLS interoperability, see the Technical White
Paper for PBB + VPLS.

25

CPE

Eth/VLAN
BGP AD H-VPLS
DSLAM

ABR

UPE

DSLAM

PW
UPE

ABR

DSLAM
ABR

UPE

DSLAM

Figure 14 H-VPLS service deployment solution

5.3 ATM DSLAM Migration

In the existing networks of many operators, the ATM DSLAM is widely
applicable to the access of individual users and enterprise private lines.
ATM switch vendors are gradually phasing out the lists of mainstream
equipment suppliers. The migration of the ATM convergence network
is necessary; there are too many access nodes in the network.
Operators cannot meet the upgrade and reconstruction costs if ATM
DSLAMs are switched to the IP DSLAM all at once. In this case, gradual
migration is more feasible; that is, IP DSLAM converging the ATM
DSLAM.
In Seamless MPLS network architecture, the IP DSLAM can directly
send services to the corresponding egress through the PWE3, without
the intermediate ATM links.
For individual user services, if the BRAS still keeps the ATM interface,
the original ATM convergence network function can be implemented
through the ATP PWE3 between the IP DSLAM and AGG node. If
the BRAS migrates to the Ethernet interface, the conversion from the
PPPOA to PPPOE and from the IPOA to IPOE is performed on the IP
DSLAM. Data is sent to the BRAS through the Ethernet PWE3.
For the enterprise private line service, the ATM PWE3 can be directly
established between the IP DSLAMs to transparently transmit ATM
services.
26

ATM

PC

RG

ATM Switch
ATM DSLAM

Business

ATM DSLAM

BRAS

PWE3 for
Residential user

PC

Business

Seamless MPLS Network

PE-AGG

IP DSLAM/OLT

RG

ATM Switch

core

metro

UPE

IP DSLAM/OLT

ATM PWE3 for Business Dedicated

Figure 15 ATM service migration scenario

27

6 Conclusion
With the simplified and unified network architecture, the Seamless
MPLS solves the O&M and reliability problems due to the isolation of
each network layer. The Seamless MPLS provides flexible and scalable
network architecture for operators.

28

7 Appendix A References
1. N. Leymann,, Seamless MPLS Architecture, draft-leymann-mplsseamless-mpls-00
2. Rosen, E., Viswanathan, A., and R. Callon, "Multiprotocol Label
Switching Architecture", RFC 3031
3. Andersson, L., Minei, I., and B. Thomas, "LDP Specification", RFC
503
4. Rekhter, Y. and E. Rosen, "Carrying Label Information in BGP-4",
RFC 3107
5. Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private Networks
(VPNs)", RFC 4364
6. Decraene, B., Le Roux, JL., and I. Minei, "LDP Extension for InterArea Label Switched Paths (LSPs)", RFC 5283
7. S. Bryant, C. Filsfils, U. Drafz, V. Kompella, J. Regan, S. Amante,
Flow Aware Transport of MPLS Pseudowires, draft-bryant-filsfilsfat-pw-03
8. D.Katz, D.Ward, Bidirectional Forwarding Detection, draft-ietf-bfdbase-11
9. K. Kompella, Y. Rekhter, Ed., Virtual Private LAN Service (VPLS)
Using BGP for Auto-Discovery and Signaling, RFC 4761
10.M. Vigoureux, D. Ward, M. Betts, Requirements for OAM in MPLS
Transport Networks, draft-ietf-mpls-tp-oam-requirements-03

29

8 Appendix B Acronyms and

Abbreviations

30

Abbreviation/Acronym

Full Spelling

MPLS

Multi-Protocol Label Switch

AN

Access Node

PE

Provider Edge

CE

Customer Edge

UPE

User-facing PE

AGG

Aggregation Node

LLU

Local Loop Unbundling

FTTC

Fiber to the Curb

FTTB

Fiber to the building

FTTH

Fiber to the home

IGP

Interior Gateway Protocol

ABR

Area Border Router

ASBR

Autonomous System Border Gateway

(LDP) DoD

Downstream on Demand

(LDP) DU

Downstream Unsolicited

FIB

Forwarding Information Base

BFD

Bidirectional Forwarding Detection

FRR

Fast Reroute

ECMP

Equal cost Multiple Path

PW

Pseudo wire

RSVP

Resource ReSerVation Protocol

VPLS

Virtual Private LAN Service

H-VPLS

Hierarchical VPLS

PBB

Provider Backbone Bridge

Copyright 2010 Huawei Technologies Co., Ltd. All rights reserved.

No part of this document may be reproduced or transmitted in any form or by any means without prior written
consent of Huawei Technologies Co., Ltd.

Trademarks and Permissions

and other Huawei trademarks are the property of Huawei Technologies Co., Ltd.
All other trademarks and trade names mentioned in this document are the property of their respective holders.

Notice
The product, service, or feature that you purchase should be restricted by
the Huawei commercial contract and the clauses in the contract. All or a
part of products, services, or features described in this document may not
be purchased or used. Every effort has been made in the preparation of this
document to ensure the accuracy of the contents, but the statements,
information, and recommendations in this document do not constitute a
warranty of any kind, expressed or implied.
The information in this document is subject to change without notice. Every
effort has been made in the preparation of this document to ensure the
accuracy of the contents, but the statements, information, and
recommendations in this document do not constitute a warranty of any
kind, expressed or implied.

Huawei Technologies Co., Ltd.

Address: Huawei Industrial Base
Bantian, Longgang
Shenzhen 518219
People's Republic of China
Website: http://www.huawei.com
Email:
[email protected]