BGP Mpls With Evpn 1677998611

BGP MPLS based EVPN KANWAR ALAM SINGH
BGP MPLS based EVPN

And its implementation and use cases
KANWAR ALAM SINGH
University of Alberta
Supervisor: Juned Noonari
Submitted to: Dr. Mike MacGregor

Abstract
MPLS based VPLS is proven L2VPN technology delivering Ethernet based Services.
However Ethernet and its requirements are continually changing due to
modernization of Networks, Ethernet Service model such as separation of Data
and Control plane, MEF based service delivery etc. In order to meet those
requirements EVPN are introduced in the industry under L2VPN IEFT Working
Group.
Intent of project is research EVPN, its requirements, its various implementation
methods and comparing it with VPLS to justify its relative benefits.
Due to EVPN being a new technology not much documentation could be found
from the vendors like Cisco and Alcatel. Juniper only being the vendor that has
the command reference for EVPN out. However, I succeeded in implementing a
Lab scenario of BGP based VPLS with Multi-homing on Alcatel Lucent 7750 SR-OS
routers.
3
Table of Contents
1.Introduction…………………………………………………………………………………..5
2.Background and Technology overview……………………………………………6
2.1 MPLS……………………………………………………………………………………………………….…..…6
2.2 L2VPNs……………………………………………………………………………………………………..….…7
2.3 VPLS………………………………………………………………………………………………………….…...8
3.Discussion………………………………………………………………………………….…22
3.1 VPLS Limitations……………………………………………………………………………………….…..22
3.2 Emerging EVPN technology……………………………………………………………………….….23
3.3 EVPN Features……………………………………………………………………………………….….….40
3.4 EVPN Applications……………………………………………………………………………….……….47
4.LAB DEMO…………………………………………………………………………………...50
5.CONCLUSION………………………………………………………………………………..54
6.BIBLIOGRAPHY……………………………………………………………………………..56
4
Table of Figures
Figure 1: A simple VLPS Model…………………………………………………………………………………9

Figure 2: Figure 2. Addition of PE router to an Existing VPLS …………………………………….12
Figure 3: Only BGP session need to set up between the new PE and RR…………..13
Figure 4. LDP VPLS ………………………………………………………………………….…………………………14
Figure 5. BGP VPLS …………………………………………………………………………………….……………..15
Figure 6. A simple VPLS topology……………………………………………………………………………….17
Figure 7.1 Logical connection between PE1-PE3, PE1-PE2 and PE2-PE4………………..…..18
Figure 7.2 Logical connection between PE1-PE3, PE1-PE2, PE2-PE4 and PE3-PE4……...19
Figure 8. EVPN……………………………………………………………………………………………………..…...24
Figure 9. EVPN-MPLS Data Plane…………………………………………………………………………….…26
Figure 10. PBB-EVPN Data Plane…………………………………………………………………………….….27
Figure 11. NVO……………………………………………………………………………………………………..……28
Figure 12 All Active Multihoming and Designated Forwarder ……………………………………40
Figure 13 All Active Multihoming and Split horizon…………………………………………………...41
Figure 14 ARP and ND………………………………………………………………………………………………..42
Figure 15 MAC mobility …………………………………………………………………………………………….43
Figure 16 Aliasing and Backup Path………………………………………………………………………..….44
Figure 17 Default Gateway………………………………………………………………………………………...45
Figure 18 MAC Mass Withdraw……………………………………………………………………………….…46
Figure 19 Layer 2 or Layer 3 Data Center Interconnect………………………………………….…..47
Figure 20 Layer 2 and Layer 3 services……………………………………………………………………….48
Figure 21 Flexible L2 and L3 Site to Site solution………………………………………………………..49
Figure 22: Network Diagram………………………………………………………………………………………51
1 Introduction
During the last decade the demand for implementing scalable L2 VPN has been
increased which led to development in the area of Layer 2 Services. Many MPLS
based technologies were brought in, Virtual Private LAN Service being one of
them. With VPLS, multipoint Ethernet services over the MPLS infrastructure can
provided using a full mesh of pseudowires. It emulates as a virtual bridge to the
end customers, and moreover they are free to run any routing protocol of their
choice. However VPLS has its own demerits, one of them being MAC learning
happens in data plane, other drawbacks include no multihoming with active-
active link, no multipoint to multipoint multicast LSPs. To overcome these
shortcomings a new solution is proposed called Ethernet Virtual Private Network.
Internet Engineering Task Force (IETF) recently published their 11th internet draft
which describes the procedures for BGP MPLS based EVPN. RFC 7209 published
by IETF describes the requirements for EVPN and how it fulfills the demands of
the new applications such Data Center Interconnect and overcomes the
limitations of current L2VPN technology VPLS.
The scope of the project is to study EVPN, specifying its advantages over VPLS and
the different implementation scenario possible.
2 Background & Technology

Overview
2.1 MPLS:
Multi Protocol Label Switching is a widely used technology since the
past decade. At first it was being used for fast switching of packets but
due to its scalability and resiliency capabilities soon it became the most
popular and vastly used technology across the networks. It simplifies
the process of routing the packets through the network by labeling the
packets, this leads to fast packet forwarding as the packet itself does
not need to be examined and the routing decisions can be made based
on the labels itself. The feature of traffic management and the
capability to support multiple service models makes it more appealing
to the ISPs.
2.2 L2VPN
Layer 2 VPN allows users to connect different LANs together over Layer
3 networks that are using MPLS or IP core. From Customer Prospective
it seems that they are connected through a big switch and feel they are
part of the same broadcast domain.
Layer 2 VPN provides users the flexibility to scale bandwidth according
to their needs. L2 Ethernet are preferred over L3 IP VPN for efficient
transport of Non-IP applications including traditional voice and video,
along with this, application security is an another L2 consideration as
the customers can control their own routing tables.
Unlike in case of Layer 3 VPN, with Layer 2 VPN service the Customer
Edge router and Provider Edge router does not need to peer up, the
customer routing tables are not stored on provider network reducing
complexity.
2.3 VPLS (Virtual Private LAN Service)
VPLS is a Layer 2 multipoint Ethernet service over the MPLS network, it

allows to connect different geographically separated LAN sites together
in a single bridged domain over the MPLS backbone and the end users
at the sites get an emulation that they are all connected on the same
LAN regardless of their actual location. Moreover customers are free to
run any Layer 3 protocol between the sites.
VPLS was originally being used to provide transparent LAN service to
enterprise customers, but now service providers are also using it as an
infrastructure technology with the emergence of Metro-Ethernet
networks.
As being a multipoint service unlike other L2VPN services, VPLS requires
just one logical interface between CE and PE to be in a full mesh. All the
PEs are fully meshed by pseudowires so that the PE receiving the frame
can identify for which VPLS instance it belongs to on the basis of
pseudowire labels. One advantage of being fully meshed is that there is
no need for the PEs to run STP to prevent loops, it uses split horizon to
prevent forwarding loops.PE maintains a separate forwarding table for
each VPLS customer.
In VPLS there is no advertisement of MAC addresses, they are always
learnt by the PEs.MAC aging mechanisms are used, the stale MAC are
removed from the forwarding table. Furthermore if the limit reaches
the MAC addresses can be removed which have not been used for a
while.
Figure 1. A simple VLPS Model
VPLS Control Plane

VPLS control plane has two main functions:
 Auto-Discovery: It refers to the process of finding out which PE
router belongs to which VPLS instance.
10
 Signaling: It Includes the setup, maintain and teardown of
Pseudowires.
There are two alternate ways of deploying control planes, one is using
LDP and the other one is using BGP. The scaling characteristics of both
these control planes differ a lot. BGP VPLS has numerous advantages
over LDP VPLS.
While using LDP, in order for the signaling of full mesh pseudowires a
full mesh of targeted LDP sessions is required between the PEs. There is
no auto discovery in case we are using LDP for signaling so these LDP
sessions must be manually configured on each PE router. These
sessions carry the VC ID which is used to identify to which VPLS
instance the LDP message refers to. While using BGP for signaling, with
auto discovery capability a PE gets to know which PEs are the members
of a given VPLS instance. The BGP NLRI contains the information that let
the automatic setting up the full mesh of pseudowires. On each PE
router a Route Target (RT) and a Route Distinguisher (RD) is configured
for each VPLS.RT is the same for a particular VPLS across all PEs and is
used to identify to which VPLS the BGP message belongs to, while RD is
used to distinguish between routes. Each PE has a identifier for each
VPLS known as VPLS Edge Identifier (VE ID), it should be unique for each
VPLS.VE ID is communicated over to other PEs.VE ID and the BGP NLRI
provides the means to the other PEs to calculate value of the pseudo
label required to reach the advertising PE. Advantage of using BGP for
signaling is that all PEs get to know about each other without any
manual configurations.
11
VPLS Forwarding Plane
Forwarding Plane Mechanisms for the unicast and to some extent for
multicast traffic are almost the same for both BGP VPLS and LDP VPLS.A
PE VPLS data plane acts as a learning bridge performing the same
functions of a standard bridge such as MAC address learning, flooding
and aging. For each VPLS instance a separate MAC forwarding table is
maintained by each participating PE router.
LDP vs BGP for control plane implementation:

Main difference between LDP and BGP schemes is that in case of LDP
there should be a full mesh of LDP sessions between PE routers while in
case of BGP a full mesh of BGP sessions is not required. This give rise to
two issues:
1. A PE router is involved in a large number of targeted LDP sessions
that leads to a limit how much a network can grow.
2. Manual configurations of LDP sessions has to be made on each PE
router whenever a new PE router is added or an existing PE router is
deleted from the network. This is definitely a cumbersome task.
12
Figure 2. Addition of PE router to an Existing VPLS requires modification on each

PE router to accomplish a full mesh of LDP sessions.
The same issues will be faced if a full mesh of BGP sessions is required,
to tackle this issue BGP Router Reflector (RR) can be used. It is in use
with L3VPN services but can be used for VPLS as well. In that case, only
RRs have to be in a full mesh. This makes it simpler when adding a new
PE router or deleting an existing one as only the BGP session between
the RR and that PE router needs to be modified.
13
Figure 3. Only BGP session need to set up between the new PE and RR.
In case of LDP VPLS, above scalability issue can be solved by

Hierarchical VPLS but at the expense of adding some other issues.
Other Difference includes the auto-discovery of the PEs, in case of LDP
there is no such mechanism so if a new PE is added to the network then
14
configurations to be changed on all the existing PE routers to make
them in full mesh with the new PE.
Figure 4. In case of LDP VPLS, manual configurations are to be made on all existing
PE routers to make them in a full mesh.
An external auto discovery mechanism can be added to the LDP VPLS.

There can be three possible combinations:
1. LDP+BGP= BGP being used only for auto discovery mechanism while
LDP being still used for signaling of pseudowires.
2. RADIUS=RADIUS server being used for auto discovery.PE sends a
request specific to VPLS instance and the server replies back with a list
of PEs belonging to that VPLS.
3. Extending LDP=No work on this topic by IETF till date.
15
While in case of BGP there is auto-discovery capability which leads to
just configuration changes on the PE to which site is attached.
Figure 5. In case of BGP VPLS, configuration is to be made on only the new PE
router rest all existing PE routers are auto discovered.
The main drawbacks of using BGP signaling:

1. Wastage of Bandwidth as Pseudowire Signaling of peer to peer
parameters are broadcast to all the PE routers.
2. Label information is broadcast to all PE routers linked with a
particular VPLS instance due to a full mesh, this is acceptable for the
initial VPLS auto discovery, but subsequent PW discovery is not
efficient.
3. BGP signaling use BGP to flush MAC address instead of IEEE Spanning
Tree TCNs making is incompatible with IEEE bridges.
(RFC 4761)
16
Multihoming
For resilience the customer maybe connected to more than one PE, in
that case forwarding loops can occur. Like in case of IP forwarding we
can use TTL to limit the circulation of a packet, in this case the customer
has to use Routers to prevent loops. If the customer chooses to use
Ethernet switch as CE then there can be great possibility of loops unless
no countermeasure is taken. One of the countermeasure is that
customer run STP on all its Switches so there is a loop free topology.
But this is not acceptable by the service provider as they have to rely on
the customer to implement it correctly, it might affect other customers
if not implemented correctly. The other countermeasure is that service
provider allow only one port to be active at a time. This can be done in
case if we use BGP for signaling, the service provider configure the
same VE ID on the PEs connecting to the same CE. By this way each PE
receives two advertisement with the same VE ID.BGP applies its
selection rules and installs the best NLRI out of them. Suppose if we
want the traffic to exit from a specific PE then we can make the local
preference higher on that PE than others, this way all PEs choose to
install the route advertised by that particular PE. In case of link failure
to that PE, other route is installed by PEs. This countermeasure is better
than relying on the customer to run STP as in this case bandwidth is
saved as broadcast traffic is sent on only one link.
Multihoming with BGP is possible by changing the local preference on

the PE but In case of LDP there is no such BGP attribute. The only way
of multihoming in LDP case is to implement H-VPLS by deploying a
17
spoke PE on the customer premises, which is homed to multiple-hub
PEs.
H-VPLS
Hierarchical version of LDP VPLS can significantly alleviate the
pseudowire scalability problem. It is achieved by reducing the number
of PE routers that have to been in full mesh, hence improving the
control plane scalability. It helps to reduce the signaling and replication
overhead to allow large scale deployments.
Taking an example, suppose the physical topology is as shown in the
figure below:
Figure 6. A simple VPLS topology.
18
All PE routers are connected to each other in a full mesh by
psuedowires.
Suppose PE1 wants to send 100mb of broadcast data then it will
replicate it three times, such that each PE router receives a copy of it.
The links PE1-P1 and P1-P2 have to carry 300mb of data which is kind of
inefficient usage of the links. With the addition of more PE routers the
amount of data which these links carry will linearly increase, hence
leading to scalability issues.
Loop problem is solved by using a split horizon rule that means a PE
router will only forward traffic to its locally attached hosts not to other
PEs, so no need for STP.
To resolve the above discussed problem, H-VPLS uses spoke
pseudowire which can carry both mesh and spoke pseudowire traffic so
that they can relay traffic between PE routers.
The above VPLS topology can be re-designed using H VPLS in two ways:
Figure 7.1 Logical connection between PE1-PE3, PE1-PE2 and PE2-PE4.
19
Figure 7.2 Logical connection between PE1-PE3, PE1-PE2, PE2-PE4 and PE3-PE4.
In the Figure 7.1, there is VPLS service defined on PE1 which has two
spoke pseudowires connecting to PE2 and PE3 and one link to local
circuit. There is no link between PE1 – PE4.
While in Figure 7.2, all the PE routers are connected in a ring by spoke
pseudowires.
In H-VPLS, split horizon rule is not followed, any traffic arriving on a
spoke pseudowire is forwarded to all other pseudowires (spoke and
meshes) and to local circuits. This leads the need to run STP (Spanning
Tree Protocol) across the network by Service Provider to avoid loops.
This can be considered one of the drawback of Using H-VLPS because
the architecture of spokes and meshes has to been planned carefully
20
else there is always chances of loops to occur. Moreover the high
convergence time of STP adds on to the problems.
H-VPLS somehow reduces the scalability issue but also adds on some
other issues.
Taking into account the topology in the Figure 7.1, Even after using the
spoke pseudowires the links PE1-P1, P1-P2 are still carrying two copies
of each broadcast traffic, hence not much of bandwidth efficiency
improvement. Second issue is the traffic reaching PE4 has to be always
relayed by PE2, as there is no direct link between PE1-PE4 leading to
congestion on PE2-PE4 link. Moreover in case of failure of PE2, PE4 get
disconnected from the other PE routers. Third issue is that PE2 router
has to learn all the MAC addresses that fall behind PE1 and PE4 because
all traffic from PE1 to PE4 will be terminated at PE2 then it is PE2 router
responsible to forward it to PE4 and vice versa.
Taking into account the other H-VPLS architecture i.e. Figure 7.2, all PE
routers are connected in a ring so there is a chance of occurrence of
loop. STP has to be run to avoid loops. Assuming if STP put the spoke
link between PE1-PE2 in block state, now there is no direct link
between PE1-PE2 this leads to traffic being relayed through PE4
creating the same issue as discussed above in case of Figure 7.1.
No doubt the advantages of H-VPLS over VPLS are:
 Improves control plane scalability by reducing the number of
pseudowires that are to be in a full mesh
 Reduces the burden on core devices by adding a hierarchical
aggregation layer.
 Size of MAC table can be reduced if combined with MAC in MAC
stacking.
21
This concludes that H-VPLS provides a solution to improve the
pseudowire scalability to some extend but not an efficient way to
handle multicast traffic. Additionally H-VPLS brings in other problems
along with it.
Operational considerations
The operational issues faced while running VPLS services are:
MAC address scaling-One of the main consideration for service provider
is the number of MAC addresses that are to be stored by every PE,
along with taking into account a PE might be providing VPLS services to
a large number of customers. Service providers have to keep an
account of the number of MAC addresses stored for each VPLS
customer and impose a limit on it by using VPLS implementations that
allow limits on per VPLS basis or per interface basis.
Limiting Broadcast and Multicast Traffic-Another concern for Service
providers is to limit the Broadcast and Multicast Traffic as it can be
costly to send such traffic especially if there are large number of PE
members and there is ingress replication. There are implementations
that allow Service providers to keep an account of this type of traffic.
Policing of VPLS traffic-Service providers need to police the amount of
traffic being send into the network by the customer on each access port
if the VPLS service is being offered over the 100 Mbps or 1 Gbps native
Ethernet ports.
22
3 Discussion
3.1 VPLS Limitations

VPLS has the following limitations that leads to the emergence of a new
technology called EVPN:
 Multihoming – Currently VPLS can only support active-standby
mode multihoming it does not support active-active redundancy
mode multihoming. Multihoming Load balancing across provider
edge nodes cannot be provided by VPLS because the MAC address
learning happens in data plane.
 Multicast optimization- No solution for leveraging Multipoint to
Multipoint LSPs with VPLS, multicast LSPs in conjunction with
VPLS is only limited to Point to Multipoint LSPs.
 Provisioning Simplicity- BGP based Auto discovery service can be
used to simplify things but still it requires the operator to
configure a number of network side parameters on top of access-
side Ethernet configuration.
 Flooding of BUM traffic – This is a big issue faced while using VPLS
services. Solution required to minimize and localise the amount of
flooding of multi-destination frames.
 Convergence Time – Network re-convergence time can be an
issue as it is dependent upon the number of MAC addresses in the
broadcast domain. Need a solution for having network re-
convergence independent of the number MAC addresses.
23
 Data Center Interconnect – VPLS cannot handle new applications
as they require the Layer 2 and Layer 3 services over the same
interface. Scalability and control like L3VPN.
In order to meet the above listed challenges, EVPN was introduced in
the industry under the L2VPN IEFT Working group.
3.2 EVPN (ETHERNET VPN):
EVPN is the next generation solution that provides L2VPN services over
the MPLS network, it inherits a decade of VPLS operational experience
in production networks but it differs from VPLS as MAC learning occurs
in control plane over the core. MAC information is carried by
multiprotocol BGP control plane and provides different choices for Data
Plane encapsulation.
EVPN is a significant milestone for the industry as it provides Layer 2
and Layer 3 service in a single VPN with scalability and control of
L3VPN.Like VPLS, EVPN allows you to connect different customer sites
located at different geographical locations together into a single virtual
LAN emulating the end users as if they are connected to the same LAN.
Now the service providers are able to meet evolving demands of higher
speeds along with sophisticated QoS. Moreover EVPN can support the
evolving demands of the new applications which cannot be met by the
existing technologies. (RFC 7209)
EVPN BASIC CONCEPTS

EVPN Terminology:
Customer Edge (CE) devices (Switch, Router and Host) connected to
Provider Edge devices.
24
EVI: It identifies a particular EVPN Instance (EVI)
ES: Set of Ethernet links connecting a multihomed device or network to
two or more PEs is known as Ethernet Segment (ES)
ESI: Ethernet Segment Identifier (ESI) is a non zero number that
identifies a particular Ethernet Segment
Ethernet Tag: It identifies a particular Broadcast or Bridge Domain in
the EVI.
Figure 8. EVPN
25
EVPN Control Plane
MAC learning over Control plane provides greater control of who learns
what, provides ability to apply policies and helps to maintain isolation
and virtualization of EVPN instances. PEs advertise MAC addresses and
IPs for next hop to other PEs over the MP-BGP with an EVPN NLRI. In
addition to load balancing over multiple LSPs between the same PEs,
control plane learning also provides load balancing of traffic between
the CEs that are multihomed to number of PEs. This helps improve
convergence time in case of one of the CE-PE link failure.
Though MAC address learning over the MPLS core happens in control
plane, but still learning between PEs and CEs is done by method
preferred by the CE, it could be data plane learning, 802.1 aq, ARP or
other protocols.
It can be locally decided whether the PE L2 forwarding table should
include all the MAC addresses known to the control plane or it should
include only the MAC addresses of the active flows or it can also
implement cache based scheme.
EVPN policy attributes are almost similar to those of L3VPN. Each EVI
needs a unique RD per PE and one or more globally unique RTs
EVPN Data Plane

EVPN abstracts and separates the control plane and the data plane.
Multiprotocol BGP as the control plane with different data plane
encapsulation choices can be as shown in figure below:
26
EVPN-MPLS Data Plane

Figure 9. EVPN-MPLS Data Plane
 EVPN-MPLS Data plane is the original EVPN solution in the base

specification, it provides a simple method of implementing EVPN
over an existing MPLS core.
 MPLS runs in the core networks control plane and data plane
providing all the MPLS features.
27
 Provides all active multihoming for Virtual Private Wire Service
(VPWS).
 It requires no pseudowires.
 IGP, RSVP-TE or LDP is required for MPLS and BGP for EVPN.
Provider Backbone Bridge (PBB) – EVPN Data Plane

Figure 10. PBB-EVPN Data Plane
 PBB-EVPN combines IEEE 802.1ah PBB with EVPN to support

scaling of very large networks with all active multihoming over
MPLS.
 It lowers the number of MAC addresses by aggregrating customer
MACs with backbone MACs, same concept as of route aggregation
in IP.
28
 Backbone Edge Bridges (BEB) PEs only advertise backbone MACs
with BGP, while customer MAC and backbone MAC mapping is
learned in the data plane.
 MPLS runs in the control plane and the data plane.
 This topology or architecture can be useful where the number of
MAC addresses are too large as this hides the customer MACs
from the backbone elevating to high MAC scalability.
Network Virtualization Overlay

Figure 11. NVO
 EVPN over NVO tunnels (VXLAN,NVGRE,MPLSoGRE) provides L2

and L3 Data Center Interconnect and resilience to simple IP
networks.
 EVPN over Virtual Extensible LAN (VXLAN) Data Plane can be used
in place of MPLS Data plane when MPLS is not available in the
core network.
29
 EVPN-VXLAN can be used to provide an L2 overlay over an IP
network. It is quite flexible, VXLAN can be routable with IP
irrespective of the underlying network being used.
 VXLAN Data plane encapsulates VXLAN header and L2 Frame
using UDP and can run over IPv4 or IPv6 while EVPN uses BGP
Control Plane to advertise MAC routes.
 Possible to provide a VPN to a hypervisor attached to a Virtual
Machine, as the VXLAN tunnel endpoints can be on Virtual
Machines.
 This architecture can provide EVPN services to DCI and virtual
network without requiring MPLS.
Ethernet Segment
If a Customer Edge (CE) is multihomed to two or more Provider Edge
(PEs), the set of Ethernet links that attach CE to PEs is known as
Ethernet Segment (ES). To CE an ES seems to be a Link Aggregation
Group (LAG). Each Ethernet Segment has a unique non zero identifier
called the Ethernet Segment Identifier (ESI). ESI is a ten octet integer in
line format sent with the most significant octet first.
The value of the ESI has to be unique and non-reserved across all the
EVPN instances on a PE. With the managed CE the ESI uniqueness
should be guaranteed by the network operator but if the CE is not
managed by the network operator then a network wide unique ESI has
to be configured for that ES. The uniqueness helps the auto discovery of
the ES and DF election.
The two reserved values of ESI are:
 ESI 0 represents a single homed CE i.e. CE is linked to only one PE.
 Maximum ESI value 0xFF (10 times) is reserved.
30
ESI has the following format:
T is a 1 octet field (most significant octet) that represents the type of

ESI. It specifies the next 9 octets i.e. the ESI value.
There are six types of ESI:
 Type 0 (T=0x00) It indicates an arbitrary ESI value which is
configured by the network operator.
 Type 1 (T=0x01) this indicates auto generated ESI value by LACP
when IEEE 802.1AX LACP is used between CEs and PEs.
 Type 2 (T=0x02) this is used in case of when there is indirectly
connected hosts (bridged LAN). The ESI value is automatically
generated by the L2 bridge protocol.
 Type 3 (T=0x03) this is MAC based ESI value that is auto generated
or configured by the network operator.
 Type 4 (T=0x04) this is Router-ID based ESI value that is auto
generated or configured by the network operator.
 Type 5 (T=0x05) this is AS based ESI value that is auto generated
or configured by the network operator.
Ethernet Tag
Ethernet Tag denotes a particular Broadcast domain such as VLAN. It is
possible to have one or more broadcast domains in a particular EVPN.
Ethernet Tag ID is a 32 bit field that contains 24 bit or 12 bit of the
identifier. The 12 bit identifier is also called the VLAN ID (VID). Service
31
providers assign VLANs to a particular EVPN. A particular VLAN may
contain more than one VID. In case of multiple VIDs within a VLAN the
participating PEs in that VLAN for a particular EVPN are responsible for
VID translation to and from the attached CEs. There is no need of VID
translation on PEs if there is only one VID in that VLAN. There are
deployment scenarios that have unique VID across all EVPN instances
and there are scenarios where all points of attachment for a particular
EVPN instance use the same VID, no other EVPN instance use that same
VID. RTs for each EVPN instance are automatically derived from the
corresponding VID.
For an EVPN instance, each PE performs a mapping of Ethernet Tag and
broadcast domain identifier (VID).
The following relationships can exist between VLANs, Ethernet Tag IDs
and MAC-VRFs:
VLAN based Service Interface:
There is only one broadcast domain (VLAN) associated with the EVPN
instance, so there is one to one mapping of VID and EVI. VID translation
is allowed, if the VLAN consists of multiple VIDs i.e. different VID per
Ethernet Segment per PE, then each PE needs to translate VID for the
frames destined to its Ethernet Segments. Overlapping of MAC
addresses across different VLANs is possible. Ethernet Tag ID in all EVPN
routes is set to zero.
VLAN Bundle Service Interface:
In this type of service interface there are multiple broadcast domains
associated with a single EVPN instance, so there is many to 1 mapping
of VID and EVI. Multiple VLANs share the same bridge domain. MAC
addresses are to be unique across VLANs. VID translation is not
allowed. Ethernet Tag ID in all EVPN routes is set to zero.
32
VLAN Aware Bundle Service Interface:
In this type of Service interface there are multiple broadcast domains
associated with a single EVPN instance with each VLAN having its own
bridge domain. In this case VID translation is allowed and Ethernet Tag
ID is assigned by Service Provider.
BGP EVPN routes:
As we know EVPN uses Control Plane for MAC learning so it introduces
a new BGP NLRI (Network Layer Reachability Information) called EVPN
NLRI.
EVPN NLRI format:
The Route type describes the Route Type Specific EVPN NLRI. While the
Length field denotes the number of octets of the Route Type Specific
field.
Multiprotocol BGP extensions AFI (Address Family Identifier) of 25 and
SAFI (Subsequent Address Family Identifier) of 70 is used to carry the
EVPN NLRI. The Two BGP speakers must use the BGP capabilities
Advertisement to properly process this EVPN NLRI.
Route types are as follows:
33
 Ethernet Auto Discovery(A-D) Route
As we already know MAC addresses are learned by BGP control plane

over MPLS core in case of EVPN, so in case of failure the convergence
time is function of MAC address advertisement routes that a PE router
needs to withdraw. This can be an issue for large network as the
number of MAC addresses would be high leading to slow convergence.
To overcome this issue EVPN introduces a mechanism using new route
types to signal the PE routers to update their forwarding tables. This
(FAST CONVERGENCE) happens by making each PE router advertise the
Ethernet Auto Discovery route type per Ethernet Segment to all remote
PE routers. When there is a connectivity failure to an attached
segment, then the PE withdraws the corresponding Ethernet Auto
Discovery route triggering each remote PE router to update their
forwarding table for the MAC addresses associated to that Ethernet
Segment and if there is any backup PE available for the same Ethernet
Segment then all the remote PE router update their next hop for all the
MAC addresses associated with that Ethernet Segment if no such
backup is there then it simply invalidates those MAC address entries for
that Particular ES.
34
Fig. Ethernet A-D route type
Fig. ESI MPLS Extended Community
ESI MPLS Extended community is also advertised along with the

Ethernet Auto Discovery Route.
The last two flags are:
R-L for Root/Leaf
A-S for Active/Standby, for single-active mode it is 1 and for all-
active mode it is 0.
35
 MAC/IP Advertisement Route
As we know the MAC address learning between CE and PE routers

happens in Data Plane. The MAC addresses behind CE are
advertised by PE router to other PEs in BGP NLRI using a MAC
Advertisement Route Type as shown in Figure below:
This Route Type is used by the PE router to advertise to remote

PEs the MAC addresses that were learned locally. MAC address
aggregation can be possible and instead of advertising each MAC
address individually, MAC prefix can be used. When MAC prefix is
36
advertised IP ADDRESS LENGTH field is set to 0 and IP ADDRESS
field is left empty. But if each MAC address is advertised
individually then IP ADDRESS field corresponds to that MAC
address. If a PE router receives an ARP request for an IP address
from a CE router, the PE router checks out the IP address and if
there is an MAC address binding for that IP address it acts as ARP
proxy and responds back to the ARP request.
MPLS Label field is of three octets, it depends on the following
procedures:
o PE router may advertise the same MPLS Label for all MAC
addresses in that particular EVI
o PE router may advertise a unique MPLS label per (ESI,
Ethernet Tag) combination
o PE router may advertise a unique MPLS label for each MAC
address
 Inclusive Multicast Ethernet Tag Route
Inclusive Multicast Ethernet Tag Route Type is advertised by a PE

router to all remote PE routers, this assists the PE router to
forward multi destination traffic to all remote PEs.
The format for Inclusive Multicast Ethernet Tag Route Type is as
shown below:
37
Suppose for a given scenario as below:
Every PE router advertises Inclusive Multicast Ethernet Tag Route

Type for that particular EVI to all other PE routers.
Suppose PE2 gets BUM packet from CE1 then it will encapsulate it
as shown:
38
After encapsulating, it will send it to PE1 only as PE3 and PE4 do

not advertise Ethernet Auto Discovery routes. The encapsulation
at PE2 occurs with adding of ESI MPLS label received from PE1 on
top of the BUM traffic and then on top of ESI MPLS label is added
Inclusive Multicast MPLS label and Transport label being finally at
the top. This route type helps with split horizon as when PE1
router receives encapsulated traffic, by checking out the ESI MPLS
Label is the same as it forwarded then it does not forward it to the
ESI.
 Ethernet Segment Route
On the basis of the information in the Ethernet Segment Route a

Designated Forwarder is elected. Ethernet Segment Route type
helps in Service carving. Service carving is the procedure of
electing multiple Designated Forwarders per Ethernet Segment
(one per EVI) to perform load balancing of multi destination traffic
destined for a segment. Election of every PE is made in a way that
it is Designated Forwarder for a disjoint set of EVIs.
39
Ethernet Segment Route format is a below:
ES Import Extended community format:
Service Carving mechanism:

o Ethernet Segment Route along with the ES import extended
community attribute is advertised by the PE router on the
discovery of ESI of the attached Ethernet Segment.
o PE router sets a timer with default value 3 seconds (This
value has to be same across all PEs connected to that ES) to
receive Ethernet Segment routes from all other PEs
connected to the same Ethernet Segment.
o With the expiry of the timer each PE creates an ordered list
of IP addresses of the PEs connected to that ES in ascending
numeric value. These IPs are obtained from the
40
ORIGINATING PEs IP ADDRESS field in the Ethernet Segment
Route. On the basis of numeric value of IP addresses an
ordinal number is assigned to each PE with 0 ordinal being
representing the IP address with lowest numeric value. The
PE router with the highest ordinal number is elected the
Designated Forwarder and the next highest ordinal number
is elected as the Backup Designated Forwarder
o The Designated Forwarder elected for that particular EVI
unblocks all traffic for the Ethernet Tags of that EVI.
In the scenario of a link/port failure, the affected PE withdraws
its Ethernet Segment Route retriggering the service carving
procedure on all PEs.
3.3 EVPN KEY FEATURES
Unlike the traditional VPLS which rely only on Data plane learning,
EVPN uses control plane for learning which adds on new features and
functions overcoming the VPLS limitations.
Following are the control plane key features:
 All Active Multihoming and Designated Forwarder
Figure 12.
41
For all active redundancy mode, the bridged network is connected to
two or more PEs using Link Aggregation Group. One of the PE is
elected as Designated Forwarder based on the information in the
Ethernet Segment Route using the Service Carving procedure. All the
BUM (Broadcast, Unknown Unicast, Multicast) traffic is forwarded by
only the DF to Ethernet Segment towards the CE as shown in the
scenario above, it avoids duplicate flooding of BUM traffic to all
active CEs. Other PEs block BUM traffic to CE.
 All Active Multihoming and Split horizon

Figure 13.
To achieve the split horizon function EVPN uses the Ethernet Auto
Discovery Route Type per Ethernet Segment, it contains the ESI label
which helps identify the Ethernet Segment this traffic originated from.
This is achieved by encapsulating every BUM traffic originating from a
non-DF PE with an MPLS label that identifies the Ethernet Segment
where the traffic originated. Egress PE use this label to filter out the
BUM traffic that has the same ESI label to where it is destined.
42
 ARP and ND
Figure 14
ARP and ND (Neighbour Discovery) flooding over the MPLS network

can be controlled by adding the associated IP address to the MAC
advertisement route type. When a PE learns the IP address
associated with a MAC address of the locally attached CE it can
advertise it to other PEs by adding it into the MAC advertisement
Route. If there are more than one IP address associated with that
MAC address then, the number of MAC advertisement route
generated are equal to the number of IP addresses. Whenever there
is an ARP request for an IP address, PE checks out it MAC address
binding for that IP address, If found it responds back to the ARP
request acting as an ARP proxy.
43
 MAC mobility
Figure 15.
With EVPN it is possible to move a host from one Ethernet segment

to another using MAC advertisement route. There would be two
MAC advertisements, one with the previous Ethernet Segment and
one with the new Ethernet Segment. Whenever a MAC move
happens all the PEs withdraw the MAC advertisement with the
previous Ethernet Segment. With local learning in data plane it is not
detectable that a MAC move has happened. Each MAC
advertisement has a sequence number with it that helps the PE
routers to update reachability to that MAC address. MAC
advertisement route with highest sequence number is saved and the
lower one is withdrawn.
44
 Aliasing and Backup Path
Figure 16.
In case of a CE multihomed to two PEs, in all active redundancy

mode using LAG, only one of the PE learns the MAC addresses of
that CE. This leads to an issue where only one PE advertises the MAC
advertisement routes for these MAC addresses to other remote PEs
making inefficient for the remote PEs to load balance traffic among
the two PEs attached to multhihomed Ethernet Segment.
EVPN comes with a solution “Aliasing”. It allows a PE to advertise
reachability to an EVI/ES even when it has learned no MAC addresses
from that EVI/ES, making possible for remote PEs to load balance
traffic among all PEs advertising it.
In case of Single Active redundancy mode, it can also be used as
backup path for that MAC address.
45
 Default Gateway
Figure 17.
Inter-subnet forwarding is possible with EVPN. PE router that

performs this is called the default gateway for that EVI, it responds
to the ARP request for the IP address that is configured as the
default gateway address. Every PE that acts as default gateway
advertises its MAC address in the EVPN control plane in the MAC
advertisement route along with the Default Gateway extended
community. The IP address field in the MAC advertisement route is
set to Default gateway IP address for that EVI, For an EVI the default
gateway IP address is same across all the PE routers in that EVI. Each
PE compares the received IP address in the MAC advertisement
route with the default gateway IP address it already contains, and if
there is any difference PE reports an error message.
46
 MAC Mass Withdraw

Figure 18.
In case of a Link failure, EVPN provides fast convergence even if

there are large number of MAC addresses associated with that. Each
PE advertises two type of routes i.e. MAC/IP address and its ESI, the
other one being connectivity to ESIs. Whenever a Failure occurs
instead of withdrawing MAC addresses individually, the PE simply
just removes the route for that ESI. Remote PEs remove the failed PE
from the path that leads to all the MAC addresses linked to the ESI.
This mass MAC removal leads to fast convergence in case of a link
failure.
47
3.4 EVPN Applications
For today EVPN can offer the following services:
 Layer 2 or Layer 3 Data Center Interconnect:

Figure 19.
o Provides scalable L2 or L3 DCI services for Virtualized Data

Centers.
o Control Plane signaling of IP/MAC provides mobility to Virtual
Machines to move between the data centers.
o Local IP gateways at each PE router optimize routing.
o Integrated L2 switching and L3 routing over the same service
interface.
48
 Layer 2 and Layer 3 services
Figure 20.
o With EVPN it is possible to provide L2 and L3 services on single

interface/VLAN to customers.
o No need for multiple VPN protocols, it provides both services
using only one network topology.
o Supports active-active or active-standby connections between
PE and CE.
o EVPN service can be run over any core network whether it be
MPLS core or IP core.
49
 Flexible L2 and L3 Site to Site solution
Figure 21.
o EVPN VXLAN can run over any IP network to provide flexible L2

and L3 VPN services to connect different sites together.
o No need of MPLS service between the sites, just needs IP
connectivity within the sites without any special configuration
to be provided by the service provider i.e. EVPN overlay is
transparent to service providers and vice-versa.
o There can be several different service providers between the
sites.
o VPN service at the endpoints can be controlled with BGP and
routing policies.
o Routing and MAC/IP advertisement is managed with IBGP
between PE routers.
50
4 Lab Demo
BGP + T-LDP based VPLS with Multi-homed CE

The lab scenario is as follows:
 3 PE routers; PE1, PE2, PE3 are running BGP based VPLS.
 1 PE router; PE4 is running T-LDP based VPLS.
 1 P router; RR is acting as Route Reflector.
 1 L2 Switch; SW1 Connected to PE1.
 1 CE router; CE2 Multihomed to PE1, PE2.
 3 CE routers; CE1, CE3, CE4 are connected to SW1, PE3, PE4
respectively.
Hardware used:
 PE1, PE2, PE3 , RR ; Alcatel Lucent 7750 SR-OS
 PE4 ; Alcatel Lucent 7710 SR-c4
 All CE ; Cisco 2911 router
 SW1 ; Cisco 3560 catalyst Switch
51
Figure 22: Network Diagram
BGP based
VPLS
SW1
192.168.0.1
50.50.50.1 192.168.0.3
50.50.50.3
CE1 CE3
10.10.10.1/30
10.10.10.10/
PE1
30
10.10.10.2/30 PE3
10.10.10.9/30 10.10.10.13/
30
192.168.0.2
T-LDP
10.10.10.6/30
CE2 RR based VPLS
50.50.50.50
MULTIHOMED
10.10.10.14/
10.10.10.5/30 30
PE2 PE4
50.50.50.2 50.50.50.4 192.168.0.4
CE4
52
Configuration steps:
 Configuring IGP: OSPF
OSPF (IGP) protocol is configured on all the PE routers as well as
on RR router for the routers to exchange their routing
information. All the routers are kept in the single domain i.e.
Area 0.
 LDP for transport tunnels

Label Distribution Protocol is required to exchange the label
mapping information. It is configured on all the core routers.
 Creating pseudo wire templates to auto create the SDPs. LDP

based pseudo wires can be automatically created using the
pseudo wire template.
 Configure BGP with address family L2VPN.

Configure BGP on PE1, PE2, PE3 and RR router with the address
family L2VPN parameter and on RR router add the cluster
command to make it the Route Reflector.
 Configure BGP VPLS on PE1, PE2, and PE3. Creating customer and
defining the Route Distinguisher and Route Target. Assigning a VE-
ID and defining the maximum value for it. The VE-ID and VE-NAME
should be unique. STP on PE1, PE2 should be shut down as
because of multihoming, it has to be run on the customer edge
router to avoid loops.
53
 PE3 and PE4 are configured to run T-LDP VPLS between them. This
way both BGP based VPLS and T-LDP based VPLS are
demonstrated.
192.168.0.3
CE3
PE3
T-LDP
based
VPLS
PE4
192.168.0.4
CE4
 Configure Multi Site on PE1, PE2 by giving same site number on

both the PE routers. SAP created on the ports that are connected
to the CE routers.
 Configure Port channel on CE2 router to make it multihomed to

routers PE1, PE2.
54
5 Conclusion
Current software codes or hardware from major vendors do not

support EVPN and its implementations. Due to non-availability of the
EVPN supported routers it was not possible to implement any of the
EVPN scenario. To conclude this report VPLS with both BGP and T-LDP
was implemented in one scenario on Alcatel Lucent 7750 SR-OS
demonstrating the limitations of VPLS. One of the biggest drawback of
VPLS is that, there is no Active-Active Multi homing support available,
one of the PE routers has to be Standby. As we know internet traffic is
rising at an enormous rate, to tackle this, load sharing with
multihoming is becoming a necessity for not only just big companies.
VPLS doesn’t support load sharing. The reason for only one link being
active is that the VPLS was developed this way. Other drawbacks
include that VPLS can’t support Multipoint to Multipoint LSPs i.e. no
optimized delivery of multi destined frames. Network re-convergence
time is dependent on the number of MAC addresses learned by the PE
55
router, this is an issue when there are large number of virtualized
machines attached at the customer end leading to a large volume of
MAC addresses to be taken care of. To demonstrate the above two
limitations, Ixia assessment tools are required. The new Data Center
Interconnect applications require Layer 2 and Layer 3 services over the
same interface added with the scalability and control like L3VPN, this is
where VPLS fails.
Future Research:
There are still many inquisitive research queries on how EVPN can will
handle the new DCI applications.
According to the authors of RFC7209 EVPN seems to overcome the
shortcomings of VPLS, but future research and practical
implementation is required to find out whether the theoretical claims
of EVPN capabilities are any true and moreover does it introduce any of
its own glitches.
One interesting research topic relevant to EVPN that comes to my mind
is that how migration from VPLS to EVPN can be done, assuming that
VPLS is already implemented and is running. Would it be possible to
move to EVPN without turning down the existing VPLS network and can
VPLS and EVPN both run together?
56
6 Bibliography
Alcatel Lucent (February 2014) ETHERNET VPN (EVPN) NEXT-GENERATION VPN FOR ETHERNET
SERVICES. Retrieved from
https://conference.apnic.net/data/37/2014-02-24-apricot-evpn-presentation_1393283550.pdf
Alcatel Lucent Scalable IP Networks Self-Study Guide by Kent Hundley (BOOK)

http://www.aldraji.com/download/ScalableIP_NRS1.pdf
Alcatel Lucent 7750 SR OS Routing Protocols Guide (book)

https://infoproducts.alcatel-lucent.com/cgi-
bin/dbaccessfilename.cgi/9300740801_V1_7750%20SR%20OS%20ROUTING%20PROTOCO.pdf
Analysis of VPLS Deployment R. Gu, J. Dong, M. Chen, Q. Zeng (Huawei) Z. Liu (China Telecom)
IETF80 L2VPN Mar. 2011 Prague
draft-gu-l2vpn-vpls-analysis-00
https://tools.ietf.org/agenda/80/slides/l2vpn-2.pdf
BGP based VPLS. Retrieved from
http://wiki.mikrotik.com/wiki/Manual:BGP_based_VPLS
BGP MPLS based Ethernet VPN

https://sites.google.com/site/amitsciscozone/home/important-tips/mpls-wiki/bgp-mpls-based-
ethernet-vpn
BGP MPLS Based Ethernet VPN draft-ietf-l2vpn-evpn-02

http://tools.ietf.org/pdf/draft-ietf-l2vpn-evpn-02.pdf
BGP MPLS Based Ethernet VPN draft-ietf-l2vpn-evpn-11

http://tools.ietf.org/html/draft-ietf-l2vpn-evpn-11
BGP Multi-Homing for VPLS Networks by Alcatel Lucent

https://infoproducts.alcatel-lucent.com/html/0_add-h-f/93-0267-
HTML/7X50_Advanced_Configuration_Guide/BGP_MH.html
57
Configuring VPLS Routing Instances by juniper
https://www.juniper.net/documentation/en_US/junos14.2/topics/usage-guidelines/vpns-configuring-
vpls-routing-instances.html
Configuring a VPLS Service with CLI by Alcatel Lucent

https://infoproducts.alcatel-lucent.com/html/0_add-h-f/93-0076-
HTML/7750_SR_OS_Services_Guide/services_vpls.html
EVPN Overview by juniper

http://www.juniper.net/techpubs/en_US/junos13.3/topics/concept/evpns-overview.html
EVPN: Intro to next gen L2VPN

http://packetpushers.net/evpn-introduction-next-generation-l2vpn/
Ethernet VPN Layer 2 Scalability by Shivlu Jain

http://www.slideshare.net/ShivluJain/ethernet-vpn-layer-2-scalability
Ethernet VPN (EVPN) for integrated layer 2-3 services – Alcatel Lucent
http://www2.alcatel-lucent.com/techzine/ethernet-vpn-evpn-integrated-layer-2-3-services/
Ethernet VPN (EVPN) and Provider Backbone Bridging-EVPN: Next Generation Solutions for MPLS-based
Ethernet Services by cisco(white paper)
http://www.cisco.com/c/en/us/products/collateral/routers/asr-9000-series-aggregation-services-
routers/whitepaper_c11-731864.html
ETHERNET VPN (EVPN) OVERLAY NETWORKS FOR ETHERNET SERVICES by Greg Hankins (Alcatel Lucent)
https://ripe68.ripe.net/presentations/170-ripe-68-evpn.pdf
Hierarchical VPLS. Retrieved from
https://sites.google.com/site/amitsciscozone/home/vpls/hierarchical-vpls
https://routingfreak.wordpress.com/2011/02/21/does-hierarchical-vpls-solve-all-scaling-issues-found-
in-vpls/
Implementing IEEE 802.1ah Provider Backbone Bridge by cisco

http://www.cisco.com/c/en/us/td/docs/routers/asr9000/software/asr9k_r4-
3/lxvpn/configuration/guide/lesc43xbook/lesc43pbb.html#90421
MPLS and VPLS security by Enno Rey.

https://www.blackhat.com/presentations/bh-europe-06/bh-eu-06-Rey-up.pdf
MPLS-Enabled Applications: Emerging Developments and New Technologies By Ina Minei, Julian Lucek
https://books.google.ca/books?id=2lxbaQ-
VN8sC&pg=PA396&lpg=PA396&dq=multihoming+drawback+MPLS&source=bl&ots=L5BUW3DvIU&sig=x
5qIErvuqKeCXlnkgM84X2zssQU&hl=en&sa=X&ei=3E7pVLiwFoyWgwSks4OIAQ&ved=0CCwQ6AEwAjgU#v
=onepage&q=multihoming%20drawback%20MPLS&f=false
58
Network Configuration Example Validating a BGP-Based VPLS Multihoming Configuration
http://www.juniper.net/techpubs/en_US/release-independent/nce/information-products/topic-
collections/nce/bgp-vpls-multihoming/validating-a-bgp-based-vpls-multihoming-configuration.pdf
(PBB-)EVPN Seamless Integration with (PBB-)VPLS draft-sajassi-bess-evpn-vpls-seamless-integ-00

https://tools.ietf.org/html/draft-sajassi-bess-evpn-vpls-seamless-integ-00
Requirements for Ethernet VPN (EVPN) (RFC 7209)

https://tools.ietf.org/html/rfc7209
Scaling Virtual Private LAN Services. Retreived from
http://blog.ine.com/2010/11/26/scaling-virtual-private-lan-services-vpls/
Scale and Extend VPLS with LDP-BGP VPLS Interworking Retrieved from
http://www.eetimes.com/document.asp?doc_id=1208544&
Scaling BGP Cisco live

http://d2zmdbbm9feqrf.cloudfront.net/2013/usa/pdf/BRKRST-3321.pdf
Technical Brief: Offering Scalable Layer 2 Services with VPLS and VLL
http://www.brocade.com/downloads/documents/technical_briefs/Offering_Scalable_Layer2_Services_
with_VPLS_and_VLL.pdf
Virtual Private LAN Service (VPLS) Interoperability with Customer Edge (CE) Bridges (RFC 6246)
Virtual Private LAN Service (VPLS) Using BGP for Auto-Discovery and Signaling(RFC 4761)
Virtual Private LAN Service (VPLS) Using Label Distribution Protocol (LDP) Signaling(RFC 4762)
Virtual Private LAN Service (Wikipedia)

http://en.wikipedia.org/wiki/Virtual_Private_LAN_Service
Virtual Private LAN Services (VPLS) by cisco

http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst6500/ios/12-
2SY/configuration/guide/sy_swcg/vpls.pdf
VPLS BGP Signaling by cisco

http://www.cisco.com/c/en/us/td/docs/ios-xml/ios/iproute_bgp/configuration/xe-3s/irg-xe-3s-
book/irg-vpls-bgp-sig.html
WHITE PAPER - LDP-BGP VPLS Interworking by juniper.

http://www.juniper.net/us/en/local/pdf/whitepapers/2000282-en.pdf

BGP Mpls With Evpn 1677998611

Uploaded by

Copyright:

Available Formats

BGP Mpls With Evpn 1677998611

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

BGP Mpls With Evpn 1677998611

Uploaded by

Copyright:

Available Formats

BGP MPLS based EVPN KANWAR ALAM SINGH

BGP MPLS based EVPN

KANWAR ALAM SINGH

Submitted to: Dr. Mike MacGregor

Figure 1: A simple VLPS Model…………………………………………………………………………………9

2 Background & Technology

2.3 VPLS (Virtual Private LAN Service)

VPLS is a Layer 2 multipoint Ethernet service over the MPLS network, it

Figure 1. A simple VLPS Model

VPLS Control Plane

LDP vs BGP for control plane implementation:

Figure 2. Addition of PE router to an Existing VPLS requires modification on each

In case of LDP VPLS, above scalability issue can be solved by

An external auto discovery mechanism can be added to the LDP VPLS.

The main drawbacks of using BGP signaling:

Multihoming with BGP is possible by changing the local preference on

3.1 VPLS Limitations

EVPN BASIC CONCEPTS

EVPN Data Plane

EVPN-MPLS Data Plane

 EVPN-MPLS Data plane is the original EVPN solution in the base

Provider Backbone Bridge (PBB) – EVPN Data Plane

 PBB-EVPN combines IEEE 802.1ah PBB with EVPN to support

Network Virtualization Overlay

 EVPN over NVO tunnels (VXLAN,NVGRE,MPLSoGRE) provides L2

T is a 1 octet field (most significant octet) that represents the type of

EVPN NLRI format:

As we already know MAC addresses are learned by BGP control plane

Fig. ESI MPLS Extended Community

ESI MPLS Extended community is also advertised along with the

 MAC/IP Advertisement Route

As we know the MAC address learning between CE and PE routers

This Route Type is used by the PE router to advertise to remote

 Inclusive Multicast Ethernet Tag Route

Inclusive Multicast Ethernet Tag Route Type is advertised by a PE

Suppose for a given scenario as below:

Every PE router advertises Inclusive Multicast Ethernet Tag Route

After encapsulating, it will send it to PE1 only as PE3 and PE4 do

 Ethernet Segment Route

On the basis of the information in the Ethernet Segment Route a

ES Import Extended community format:

Service Carving mechanism:

 All Active Multihoming and Split horizon

ARP and ND (Neighbour Discovery) flooding over the MPLS network

With EVPN it is possible to move a host from one Ethernet segment

In case of a CE multihomed to two PEs, in all active redundancy

Inter-subnet forwarding is possible with EVPN. PE router that

 MAC Mass Withdraw

In case of a Link failure, EVPN provides fast convergence even if

 Layer 2 or Layer 3 Data Center Interconnect:

o Provides scalable L2 or L3 DCI services for Virtualized Data

o With EVPN it is possible to provide L2 and L3 services on single

o EVPN VXLAN can run over any IP network to provide flexible L2

BGP + T-LDP based VPLS with Multi-homed CE

Figure 22: Network Diagram

 LDP for transport tunnels

 Creating pseudo wire templates to auto create the SDPs. LDP

 Configure BGP with address family L2VPN.

 Configure Multi Site on PE1, PE2 by giving same site number on