Global vision.

Local knowledge.
Cisco Connect Dubrovnik
27.-29.3.2019.
EVPN in Service Provider network

Dejan Jaksic
Sytstems Engineer Service Providers
28-3-2019
 • Planet EVPN motivation
• EVPN Basics
• Network Fabric Architecture
Agenda • EVPN-VPNv4 interconnect
• EVPN and VPLS seamless
integration
• EVPN positioning in SP network
• Conclusion
EVPN: Value Proposition

Create New Revenue Deploy with Ease

Streams • Seamless Brownfield Integration
• Same principles and operational
• Network as a service through fabric
experience as IP VPNs
designs
• E-LAN, E-LINE, E-TREE, L3, IRB
Services EVPN
Increase Availability
Protect Investments • Workload Mobility
• Unified Networks on single overlay • Optimal forwarding
• Simplify protocols and operations • All-Active Redundancy with Fast
• Industry adoption and standardization Convergence

Fast, Resilient, Flexible Unified Services

 EVPN Unified Services Attributes
Seamless
Deployment

Access Services Optimal East-

West traffic
delivery

All-Active Multi-
Homing
EVPN Per-Flow
Redundancy
and load-
balancing

Workload Mobility
Fast
Convergence

EVPN external Hub: https://e-vpn.io/

Evolution of Ethernet

Ethernet IEEE IETF IETF

II 802.1Q VPLS TRILL

1973 1982 1983 1998 2006 2007 2008 2011 2015

Ethernet IEEE
Ethernet IEEE IETF
Over 802.1ah
Begins 802.3 EVPN
MPLS (PBB)
Stolen Data Center requirements J

§ Flexible service/workload placement

§ Multi-tenancy with L2 and L3 VPN
§ Optimal Forwarding, Workload mobility
§ Fast Convergence
§ Efficient bandwidth utilization

• EVPN with a choice of data plane encapsulation (MPLS/SR, VxLAN, SRv6) is

the designed technology to address these requirements.
What is EVPN?
RFC 7432
EVPN family introduces next EVPN
generation solutions for Ethernet
services
P2P Multipoint
• BGP control-plane for Ethernet Segment
and MAC distribution learning over MPLS
and VXLAN data-plane
• Same principles and operational EVPN-VPWS
experience as in IP VPNs
RFC 8214
No use of Pseudowires
Multi-vendor solutions EVPN-IRB
RFC 7432
EVPN
Cisco leader in industry
standardization efforts (RFCs/Drafts)
draft-ietf-bess-evpn-inter-subnet-forwarding
Concepts
EVPN Instance (EVI) Ethernet Segment BGP Routes BGP Route Attributes

SHD CE1
Route Types Extended Communities
BD EVI ESI1 PE1 [1] Ethernet Auto-Discovery (AD) Route ESI MPLS Label

[2] MAC/IP Advertisement Route ES-Import

MHD CE2
[3] Inclusive Multicast Route MAC Mobility
BD
EVI

PE2
ESI2 [4] Ethernet Segment Route Default Gateway
PE
[5] IP Prefix Advertisement Route Encapsulation

• EVI identifies a VPN in the
network
• Encompass one or more
bridge-domains,
depending on service
interface type
Port-based
VLAN-based (shown above)
VLAN-bundling
• Represents a ‘site’
connected to one or more
PEs
• Uniquely identified by a 10-
byte global Ethernet
Segment Identifier (ESI)
• Could be a single device
or an entire network
Single-Homed Device (SHD)
Multi-Homed Device (MHD)
Single-Homed Network (SHN)
Multi-Homed Network (MHN)
• New SAFI [70]
• Routes serve control
plane purposes,
including:
MAC address reachability
MAC mass withdrawal
Split-Horizon label adv.
Aliasing
Multicast endpoint discovery
Redundancy group discovery
Designated forwarder election
IP address reachability
L2/L3 Integration
• New BGP extended
communities defined
• Expand information
carried in BGP routes,
including:
MAC address moves
Redundancy mode
MAC / IP bindings of a GW
Split-horizon label encoding
Data plane Encapsulation
Service Provider Network - Simplification Journey

Compass
Unified MPLS EPN 5.0 Metro Fabric
Provisioning NETCONF NETCONF
YANG YANG

Programmability

L2/L3VPN Services LDP BGP LDP BGP BGP

Inter-Domain CP BGP-LU BGP-LU

FRR or TE RSVP
IGP with SR
LDP IGP with SR
Intra-Domain CP
IGP

https://xrdocs.io/design/
 Drastic Network Protocols Reduction

Bag of existing Protocols

Next Gen.
Ethernet MP-BGP
802.1Q,
Protocols
LDP
802.1ad LDP-TE SRv6 Key enabler for
IPv4 IP OAM SR (MPLS)
PPPoE Reducing operations complexity
MPLS OAM PCEP
IPv6 Ethernet OAM § Simpler automation
ISIS
MPLS STP BGP (TE, LS) § Simpler to repair
L2TP G.8032 IP OAM § Simpler integration
PWE3 RADIUS Ethernet OAM § Foundation for service Orchestration
ISIS SNMP EVPN
OSPF Syslog NETCONF/YANG
RSVP-TE Netflow SSH
LACP SSH CLI/XML
MC-LACP HSRP/VRRP
Why was EVPN needed?
Solving VPLS challenges for per-flow Redundancy

• Existing VPLS solutions do not offer an All- M1 M2

Active per-flow redundancy – VPLS CE1 PE1 PE3 CE2
technology lacks the capability of Echo !
preventing L2 loops PE2 PE4

• Looping of Traffic Flooded from PE (BUM)

M1 Duplicate !
M2
• Duplicate Frames from Floods from the PE1 PE3 CE2
CE1
Core
• MAC Flip-Flopping over Pseudowire PE2 PE4

• E.g. Port-Channel Load-Balancing does not

produce a consistent hash-value for a M1 M2
frame with the same source MAC (e.g. non CE1 PE1
MAC
Flip-Flop PE3 CE2
MAC based Hash-Schemes)
PE2 PE4
Why was EVPN needed in 2016+?

• Network Operators have emerging needs in their network:

• Data center interconnect operation (DCI)
• Cloud and Services virtualization (DC)

• Remove protocols and Network Simplification (ICCP, HSRP/VRRP)

• Integration of Layer 2 and Layer 3 VPN Services

• What about PBB-EVPN?

• ASR9k only and slowing down investments

EVPN is all about BGP Control Plane and Services!!!

From MAC Bridging to MAC Routing
Common BGP Control Plane
L3VPN – VPNv4/6, L2/3VPN – EVPN, EVPN-VPWS Overlay
Evolution:
Underlay
IP, MPLS IP,MPLS,VXLAN IP,MPLS,VXLAN

Data Center Network

Service Provider Network overlap

Leaf
VM

PE1 DCI1

Spine Spine

Leaf
VM
A1 Access WAN/Core

Leaf
PE2 DCI2 VM

Existing Solution: L2/L3VPN (BGP,T-LDP) - VPLS, EoMPLS VPLS, OTV Trill, Fabric-Path Overlay

IP, IGP, MPLS (LDP), RSVP-TE, BGP-LU IP, MPLS, L2 L2, STP, VLAN Underlay
 MPLS Transport & BGP Service
BGP L3VPN BGP EVPN
BGP Signaling BGP Signaling BGP Signaling BGP Signaling

PE2 PE4 PE2 PE4

CE1 MPLS CE2 CE1 MPLS CE2

PE1 PE3 PE1 PE3

Data Plane Data Plane

IP Packet Transport IP Packet L2 Frame Transport L2 Frame

MPLS Label MPLS Label
Service Service
BGP Label BGP Label
IP Packet L2 Frame
EVPN - Ethernet VPN
• Concepts are same!!! Pick your side!

Pick your side!

SP1 SP2
PE2 PE4
CE1
PE1 PE3
L1 L2 L3 L4

C1 C2
VM VM VM VM
EVPN vs VPNv4/6 or BGP Control Plane?
• BGP integrates services with programmable SR transport
• Common across L2 / L3 services
• Services Control Plane is BGP with different AF / SAFI
• Single Service Control Plane is easy to manage and troubleshoot
• HUGE investment in existing VPNv4/6
• EVPN doesn’t replace L3VPN VPNv4/6 - no technical reason to do it!
EVPN Flavors
• Multi-Homed All-Active Ethernet Access
• Replacement of: mLACP, STP, T-LDP, BGP-AD, etc.
• Standards-based Multi-chassis / Cluster Control Plane
• Replacement of: vPC, VSS, nVCluster, etc.
• Replacement of: HSRP, VRRP, etc.
• Carrier Ethernet Today
• E-LINE - 80% of SP’s L2VPN portfolio (PWs)
• E-LAN - Smaller # of L2 Multipoint VPN services
• There isn’t other standard technology with Ethernet All-Active Multi-
homing
EVPN - Positioning
• EVPN should be door-opener for IOS XR in Next Generation CO
(Network Fabric)
• EVPN L2/L3 multipoint brings optimal forwarding, MAC mobility, all-active
MH access
• BGP Control-Plane
• Simplifies DCI/Border-Leaf configuration and service provisioning
• Provides common SLA signaling
EVPN - Ethernet VPN
• Leafs run Multi-Protocol BGP to advertise & learn MAC/IP addresses over the
Network Fabric
• MAC/IP addresses are advertised to rest of Leafs

SP SP

MAC/IP advertisement &

learning via BGP EVPN NLRI

L L L L
Data Plane learning
from the hosts
All Active multi-homing
C C Ethernet Segment
VM VM VM VM
XR CLI:
Step 2:
Step 1:
EVPN - Ethernet-Segment for Multi-Homing

SP1 SP2
The bundle on the Leafs
connecting to a node should
have Identical ES identifier (ESI)

L1 L2 L3 L4
Unique 10-byte global identifier
per Ethernet Segment Ethernet Segment represents a
C1 C2 node connected multiple Leafs

VM VM VM VM
EVPN – Designated Forwarder (DF)
Challenge:
How to prevent duplicate copies of flooded traffic from being delivered to a multi-homed
Ethernet Segment (BUM traffic)?

SP1 SP2

L1 L2 L3 L4
NDF DF
C1 Duplicate C2
EVPN – Split Horizon
Challenge:
How to prevent flooded traffic from echoing back to a multi-homed Ethernet Segment?

Transport
BUM Label Label

SP1 SP2
SH Label

L1 L2

C1 Echo !

VM VM
EVPN – MAC Mass-Withdraw
Challenge:
How to inform other Leafs of a failure affecting many MAC addresses quickly while the
control-plane re-converges?

SP2 MAC1 à ESI1 à Leaf1 + Leaf2

MAC1 can be SP1
reached via ESI1

L1 L2 L3 L4
MAC1 can NOT be
reached via ESI1
C1 C2
VM VM VM VM

ESI1 MAC1
 BUM = Broadcast, Unknown unicast, Multicast

EVPN – BUM Ingress Replication

SP1 SP2
BU

BU
BU
M

M
M

L1 L2 L3 L4
BU
M

C1 C2
VM VM VM VM
R36, R37, R38, R39 - EVPN Startup
R36 - Example
1. RT4: DF Election & Multi-Homed Ethernet
Segment Auto-Discovery
LACP R39
Service Carving: 100 modulo 2 = 0 H2
R36 is DF for EVI-100
R38 R35

RT-4 - DF Election

LACP R37 R34 RD: 1.1.1.36:1

H1 ESI: 0036.3700.0000.0000.1100

R36 Ext-Com: 3637.0000.0000 (RT)

R36, R37, R38, R39 - EVPN Startup
R36 - Example
1. RT4: DF Election & Multi-Homed Ethernet
Segment Auto-Discovery
LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery
(Split-Horizon, Mass-Withdraw) H2
RT-1 - Per ESI Ethernet AD
R38 R35
RD: 1.1.1.36:1

ESI: 0036.3700.0000.0000.1100

LACP R37 R34 Flag:0x00 All-Active

Ext-Com:
Split-Horizon Label: 64005
H1 Ext-Com: 1:100 (RT)

R36
R36, R37, R38, R39 - EVPN Startup
R36 - Example
1. RT4: DF Election & Multi-Homed Ethernet
Segment Auto-Discovery
LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery
(Split-Horizon, Mass-Withdraw) H2
3. RT3: Inclusive Multicast RT-3 - Inclusive Multicast
R38 R35
RD: 1.1.1.36:100

Ext-Com: Type 6 Ingress-Replication

Multicast(BUM) Label: 64120
LACP R37 R34
Ext-Com: 1:100 (RT)

H1
R36
BUM Forwarding

1. RT4: DF Election & Multi-Homed Ethernet

Segment Auto-Discovery
LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery
(Split-Horizon, Mass-Withdraw) H2
3. RT3: Inclusive Multicast
R38 R35

LACP R37 R34

H1 Transport Label R38-9

R36 BUM Label R38-9/EVI100

BUM - Traffic
IR BUM - Traffic
BUM Forwarding

1. RT4: DF Election & Multi-Homed Ethernet

Segment Auto-Discovery
LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery
(Split-Horizon, Mass-Withdraw) H2
3. RT3: Inclusive Multicast
R38 R35

X
LACP R37 R34
Transport Label R37
H1 BUM Label R37/EVI100

R36 SH Label R37/ESIx

BUM - Traffic
IR BUM - Traffic
R36, R37, R38, R39 - EVPN Startup
R36 - Example
1. RT4: DF Election & Multi-Homed Ethernet
Segment Auto-Discovery
LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery
(Split-Horizon, Mass-Withdraw) H2
3. RT3: Inclusive Multicast RT-2 - MAC Advertisement
4. RT2: MAC Advertisement R38 R35
RD: 1.1.1.36:100

ESI: 0036.3700.0000.0000.1100

LACP R37 R34

MAC: 0062.ec71.fbd7

H1 Label: 64004

R36 Ext-Com: 1:100 (RT)

L2 Frame SMAC:
0062.ec71.fbd7

per EVPN Bridge Domain label

Unicast Forwarding (looks familiar?)
L2 Frame Flow1 Transport Label R36
1. RT4: DF Election & Multi-Homed Ethernet DMAC: H1

Segment Auto-Discovery RT-2 MAC Label/EVI

LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery L2 Frame Flow1
DMAC: H1
(Split-Horizon, Mass-Withdraw) H2
3. RT3: Inclusive Multicast
4. RT2: MAC Advertisement R38 R35

LACP R37 R34

H1
R36
L2 Frame Flow1
DMAC: H1
EVPN Routes – Cheat Sheet
PE1 – Advertises:
BGP Signaling

RT-4 Ethernet Segment Route

PE2 PE4 PE1 • I have ESI1 in case when someone needs this information for
CE1 MPLS CE2 Designated Forwarder(DF) Election
EVI1-L
PE1 PE3
BE1-SHL RT-1 Per ESI Ethernet Auto-Discovery (AD) Route
Data Plane
L2 Frame Transport L2 Frame
EVI1-BUML • I have ESI1
MPLS Label
Service
• ESI1 is All-Active
BGP Label
BD1 EVI1 • AC with ESI1 is connected to EVI1 and EVI2
L2 Frame
• My Split Horizon Label for ESI1 is
BD1 MAC BE1-SHL
MAC-A

.1
BVI1
MAC-A -> BE1.1
RT-1 Per EVI Ethernet Auto-Discovery (AD) Route(s)

BE1
IP-A
BE1 - ESI1
VRF1 ARP • EVI1 per-EVI (Aliasing) Label is
Vlan1 IP-A MAC-A -> BVI1
• EVI2 per-EVI (Aliasing) Label is EVI1-L
VRF1 IP-B MAC-B -> BVI2
EVI2-L
Vlan2 VRF1-AGGL RT-3 Inclusive Multicast Route(s)
BE1

MAC-B BVI2 • EVI1 Label for BUM traffic is

EVI1-BUML
BD2 MAC • EVI2 Label for BUM traffic is
.2

IP-B MAC-B -> BE1.2 EVI2-BUML

BD2 EVI2 RT-2 MAC/IP Advertisement Route(s)

• MAC-A/IP-A in EVI1 and IP-A in VRF1 via label
EVI1-L
EVI2-L
• MAC-B/IP-B in EVI2 and IP-B in VRF1 via label
EVI2-L
BE1-SHL
RT-5 Prefix Advertisement Route(s)
EVI2-BUML
• IPv4/6 prefix of BVI1 in VRF1 via label VRF1-AGGL
• IPv4/6 prefix of BVI2 in VRF1 via label VRF1-AGGL
 EVPN – Distributed Anycast Gateway
Purpose:
Optimal intra and inter-subnet connectivity with seamless workload mobility
Identical Anycast Gateway Virtual IP
and MAC address are configured on
all the Leafs
Distributed Anycast Gateway serves
as the gateway for connected hosts SP1 SP2

BVI BVI BVI BVI

GW GW GW GW

L1 L2 L3 L4
All the BVIs perform active forwarding
in contrast to active/standby like First-
hop routing protocol
C1 C2 C3 C4
VM VM VM VM
EVPN – IRB in Network Fabric
CORE Routing

Intra-subnet
Forwarding

SP SP
Inter-subnet
Forwarding L3
GW GW GW GW --
L2
L L L L
Subnet 1

Subnet 2 C C C C
VM VM VM VM

40
 Centralized vs. Distributed Routing

Distributed Routing (IOS XR) Centralized Routing

Boarder L3 Centralized GW
Leaf
L2

Fabric Fabric
L3
Leaf Leaf
L2

Subnet 1 Subnet 2 VLAN 1 VLAN 2

• Optimized forwarding of east-west traffic • All east<->west routed traffic traverses to centralized gateways
• ARP/MAC state localized to Leafs • Centralized gateways have full ARP/MAC state in the DC
• Helps with horizontal scaling of DC • Scale challenge
• We do NOT support this design!
 Integrated Routing and Bridging
Symmetric IRB Asymmetric IRB

Boarder Boarder
Leaf Leaf

Fabric Fabric
Leaf Leaf

• Flexible workload placement – any subnet

anywhere
• Bridge->Route/Route->Bridge (symmetric VNI
in both directions)
• ARP/MAC state localized to Leafs
• Helps with horizontal scaling of DC
• Cisco supports ONLY this mode
• Egress subnet must be local
• Bridge->Route->Bridge (Different (Asymmetric) VNI
depending on directions)
• Ingress Leaf needs ARP/MAC state for every egress
leaf
• Limits scale

Symmetric IRB and Asymmetric IRB are NOT interoperable!

LR36, R37, R38, R39 - EVPN Startup
R36 – Example L2/L3 service
Anycast IRB 192.168.2.1/24

1. RT4: DF Election & Multi-Homed Ethernet IRB

Segment Auto-Discovery
LACP R39
2. RT1: Per ESI Ethernet Auto-Discovery
(Split-Horizon, Mass-Withdraw) H2 IRB
3. RT3: Inclusive Multicast RT-2 - MAC Advertisement
4. RT2: MAC/IP Advertisement R38 R35
RD: 1.1.1.36:100

IRB ESI: 0036.3700.0000.0000.1100

LACP R37 R34

MAC: 0062.ec71.fbd7

H1 Label: 64004(BD) + 64008(VRF)

R36 IP: 192.168.1.10

L2 Frame SMAC: Ext-Com: 1:100 (RT) + VRF RT

0062.ec71.fbd7
IP Header SurceIP:
192.168.1.10
Anycast IRB 192.168.1.1/24
 EVPN - load-balancing modes
All-Active Single-Active Port-Active
(per flow) (per VLAN) (per port)

PE1 PE2 PE1 PE2 PE1 PE2

V1 V1 V1 V2 V1, V2

CE CE CE

Single LAG at the CE Multiple LAGs at the CE Multiple LAGs at the CE

VLAN goes to both PE VLAN active on single PE Port active on single PE
Traffic hashed per flow Traffic hashed per VLAN Traffic hashed per port
Benefits: Bandwidth, Convergence Benefits: QoS, Billing, Policing Benefits: Protocol Simplification
Can replace: vPC, VSS, nV cluster Can replace: HSRP/VRRP Can replace: ICCP MC-LAG
 EVPN - load-balancing modes
Single-Flow-Active
(access L2 GW)

PE1 PE2

H1 H2

CE1 CE2 CE3

Single LAG at the CE

VLAN goes to both PE
Access takes care of L2 loop
Benefits: Legacy support for STP,
REP, G.8032
EVPN-VPWS
• Benefits of EVPN applied to point-to-point services
• No signaling of PWs. Instead signals MP2P LSPs instead
(like L3VPN)
• All-active CE multi-homing (per-flow LB) - today
• Single-active CE multi-homing (per-service LB) - roadmap
PE PE
• Relies on a sub-set of EVPN routes to advertise CE CE
Ethernet Segment and AC reachability
• PE discovery & signaling via a single protocol – BGP PE PE
• Per-EVI Ethernet Auto-Discovery route (RT1 only)!!!
EVPN – Flexible Cross-Connect Service
Challenge:
How to bring multiple access services from different sources using a single EVPN E-LINE
tunnel?
A1
CE1 CE1
A1
EVPN
PE CE2 PE
CE2 Forwarder
MUX
A2 CEn DEMUX
CEn

Normalized VLAN
VLAN translation over unique tunnel

Single MPLS label used

Flexible Cross-Connect Service: Head-End
Purpose:
Bring multiple ELINE services into Pseudo-wire Head-end termination
PWHE
CE1
A1 PE1

CE2

A2 PE2
CEn
PWHE

ELINE
(single tunnel)
VLANs VRFs
EVPN - L3 Multi-Homing using EVLAG
No ICCP!
HSRP/VRRP/MC-LAG Replacement
Access Core / Metro Fabric Access

BVI
CE1 GW PE1 PE3 CE3

L3VPN
or
EVPN
BVI
CE2 GW PE2 PE4 CE4

EVLAG EVLAG

EVPN Multi-homing is provided by EVPN (EVLAG)

Determine DF/ NDF PE
Synchronization (ARP, IGMP, etc.)
EVPN - Anycast-PW

Access Core / Metro Fabric Access

AG AG

CE A1 EVPN A3 CE
Anycast-PW All-Active Anycast-PW

AG AG

Anycast-SID Anycast-SID
EVPN - access VPWS (H-EVPN)
Multi-Homed EVPN-VPWS - Roadmap
Access Core / Metro Fabric Access

A1 AG AG A3

EVPN-VPWS EVPN EVPN-VPWS

CE CE

A2 AG AG A4
EVPN and VPNv4/6 Interconnect
• DCI/BL provides EVPN to VPNv4/6 stitching
• DCI/BL participates in L3 Routing not in L2 Bridging
• DCI/BL is mandatory, because of summarization!!!

BGP - L3VPN VPNv4/6 BGP - EVPN

LEAF
A P S

CE2 A Access ABR Core DCI/BL DC/CO CE1

A P S

LEAF
EVPN and VPNv4/6 Interconnect
• DCI/BL provides EVPN to VPNv4/6 stitching Interconnect
• DCI/BL participates in L3 Routing not in L2 Bridging
• DCI/BL is mandatory, because of summarization!!!

BGP - L3VPN VPNv4/6 BGP - EVPN

RT5 Prefix = prefix-CE2/24 RT5 Prefix = prefix-CE2/24

RT5 Prefix = prefix-CE1/24 RT5 Prefix = prefix-CE1/24

RT2 MAC/IP = CE1/32

X

LEAF
A P S

CE2 A Access ABR Core DCI/BL CO CE1

A P S

LEAF
EVPN and VPNv4/6 Interconnect

BGP - L3VPN VPNv4/6 BGP - EVPN

RT5 Prefix = prefix-CE2/24 RT: VRF A RT5 Prefix = prefix-CE2/24 RT: VRF A Stitching

RT5 Prefix = prefix-CE1/24 RT: VRF A RT5 Prefix = prefix-CE1/24 RT: VRF A Stitching

RT2 MAC/IP = CE1/32 RT: VRF A Stitching

X
DCI/BL

VRF A
RD DCI:0
RT import/export: VRF A Stitching
RT import/export: VRF A
 EVPN and VPNv4/6 Interconnect
EVPN to VPNv4/6 Re-Advertise

BGP - L3VPN VPNv4/6 BGP - EVPN

RT5 Prefix = prefix-CE1/24 RT: VRF A RT5 Prefix = prefix-CE1/24 RT: VRF A Stitching

RT2 MAC/IP = CE1/32 RT: VRF A Stitching

1. Import: RT: VRF A Stitching
router bgp 1
3. Filter RT2 => /32 Router
2. Advertise to vpnv4: VRF A
DCI/BL - BGP Configuration X
address-family l2vpn evpn DCI/BL
import stitching-rt re-originate
advertise vpnv4 unicast re-originated stitching-rt
! VRF A
address-family vpnv4 unicast
RD DCI:0
import re-originate stitching-rt
route-policy rt2-filter out RT import/export: VRF A Stitching
advertise vpnv4 unicast re-originated RT import/export: VRF A
!
 EVPN and VPNv4/6 Interconnect
VPNv4/6 to EVPN Re-Advertise

BGP - L3VPN VPNv4/6 BGP - EVPN

RT5 Prefix = prefix-CE2/24 RT: VRF A RT5 Prefix = prefix-CE2/24 RT: VRF A Stitching

2. Advertise to EVPN: RT: VRF A Stitching

DCI/BL - BGP Configuration
router bgp 1
address-family l2vpn evpn DCI/BL
import stitching-rt re-originate
advertise vpnv4 unicast re-originated stitching-rt
! VRF A
address-family vpnv4 unicast
RD DCI:0
import re-originate stitching-rt
route-policy rt2-filter out RT import/export: VRF A Stitching
advertise vpnv4 unicast re-originated RT import/export: VRF A
1. Import: VRF A
!
 VPLS & EVPN Seamless Integration - Migration

VFI1 is by default in Split Horizon Group 1

CE2 PE2 PE4 CE4
• SHG1 protects loops in MPLS Core
• Full Mesh of pseudowires (PW) is required
for Any-to-Any forwarding
PE1 MPLS
VFI1
PW_PE2 UP
BD1 PW_PE3 UP
PW_PE4 UP
CE1
PE3 CE3

Migrate VPLS Network to EVPN Network through Seamless Integration

 VPLS & EVPN Seamless Integration - Migration

VFI1 is by default in Split Horizon Group 1

CE2 PE2 PE4 CE4
• SHG1 protects loops in MPLS Core
• Full Mesh of pseudowires(PW) is required
for Any-to-Any forwarding
PE1 MPLS
VFI1 EVI1 is also by default in Split Horizon Group 1
PW_PE2 UP
BD1 PW_PE3 UP • PE1 doesn’t forward data between VFI1 and EVI1
CE1 SHG1 X PW_PE4 UP

EVI1 PE3 CE3

 VPLS & EVPN Seamless Integration - Migration

VFI1 is by default in Split Horizon Group 1

CE2 PE2 PE4 CE4
• SHG1 protects loops in MPLS Core
• Full Mesh of pseudowires(PW) is required
for Any-to-Any forwarding
PE1 MPLS
VFI1 EVI1 is also by default in Split Horizon Group 1
PW_PE2 UP
BD1 PW_PE3 DOWN • PE1 doesn’t forward data between VFI1 and EVI1
SHG1 X PW_PE4 UP
CE1 BGP EVPN
EVI1 PE3 CE3 PE1&PE3 run BGP EVPN
• PW_PE3 goes DOWN
• Data Forwarding between PE1 and PE3 via EVI1
EVPN – MVPN in the Network Fabric (in progress)
XR 6.6.1 - IGMP L2 EVPN state sync

Source
VRF(x)

BL BL
MVPN State sync
in EVPN
SP SP

L3
---
L2 L L L L
EVI-x
IGMP Join / Leave
EVI-y
C C
IRB vrf(x)
mcast Receiver Receiver
evpn
EVPN – Service Layering
Access Aggregation Core
A AG PE P PE
Multicast CE
FXC
EVPN-HE
E-TREE A AG PE P PE
P2P
L2 Bridging
L3 Routing BL BL
IRB
SP SP
SP SP
SP SP
SP SP “Shared or single tenant”
L L L
L L L
L L L
L L L L
L L L L
L L L L

C C CE CE
 Service Provider Network BGP/EVPN

Metro Core Metro

A A AG PE/P P PE/P AG A A

CE CE

A A AG PE/P P PE/P AG A A

o Distributed Anycast Gateway EVPN-IRB

o All-active Multi-Homing
o EVPN symmetric IRB
o EVPN L2 for east-west traffic
o EVPN L3 for north-south traffic

ü Seamless mobility
ü Optimal forwarding (east-west & north-south)
ü All-active multi-homing load-balancing
ü Allow virtualization of appliances
ü EVPN multi-services (E-LAN, E-LINE, IRB)
ü Optimal bandwidth utilization within fabric
 SP Routing Use Case Representation

PRE-AGG DCI
5G

Mobile FH Direct
MOBILE EDGE Content
CELL SITE ROUTER METRO AGG
4G H
SP PEERING
B
bile
Mo

B2B EDGE SPINE

Enterprise IP or MPLS CORE
Internet
LEAF
BUS ACCESS
TOR
Residential CABLE AGG RESIDENTIAL EDGE

Public Cloud
CIN CABLE EDGE
CLOUD PEERING
RPD

Access Aggregation Edge Network Fabric Core Peering

EVPN Applicability
EVPN Advantages:
Integrated • Integrated Layer 2 and Layer 3 VPN services

Services • L3VPN-like principals and operational experience for scalability and control
• All-active Multi-homing & PE load-balancing (ECMP)
• Fast convergence (link, node, MAC moves)
Network
• Control-Place (BGP) learning. PWs are no longer used.
Efficiency
• Optimized Broadcast, Unknown-unicast, Multicast traffic delivery

• Choice of MPLS, VxLAN or SRv6 data plane encapsulation

Service • Support existing and new services types (E-LAN, E-Line, E-TREE)
Flexibility
• Peer PE auto-discovery. Redundancy group auto-sensing

Fully support IPv4 and IPv6 in the data plane and control plane
Investment •

Protection • Open-Standard and Multi-vendor support

Conclusion
• EVPN is an very important complement to BGP based services
• BGP is Unified Services Control Plane across SP Network
• EVPN All-Active Multihomed Service with Distributed Anycast Gateway & Integration
to L3VPN simplifies SPDC/NextGen-MEC/WAN Integration
• Cisco 5G xHaul vision: Segment Routing for underlay data plane + BGP EVPN for
control plane end to end!

Cisco xHaul whitepaper

EVPN in Service Provider network