SEGMENT ROUTING

FOR SDN
Shaowen Ma, APAC Product Director, Juniper, [email protected]
March 1, 2017

JUNIPER CONFIDENTIAL
AGENDA

Introduction
Segment Routing Deep Dive
Segment Routing SDN and Use Case
Summary

Juniper Confidential
 MPLS – 16 YEARS, GREAT SUCCESS
THE ACTUAL STANDARD FOR SERVICE DELIVERY
• LDP, mLDP

• RSVP-TE, RSVP-TE P2MP

• L3 MPLS VPN

• 6VPE/6PE

• L2 MPLS VPN – VPWS

• L2 MPLS VPN – VPLS (LDP, BGP, BGP AD)

• Next-generation multicast VPN

Kireeti Kompella Eric Rosen Yakov Rekhter Many…
• MPLS-OAM, LSP BFD, VCCV Ping, and VCCV-BFD

• MPLS-TP Static LSP/PW, OAM, APS IETF SPRING/Segment Routing working group
• Source Packet Routing in Networking
• GMPLS, GMPLS UNI*
SDN 2.0 ERA

SR
Controller
Controller
OpenFlow Segment Routing

PE1 ASBR

Segment Routing, RSVP-TE Enable SDN 2.0

Edge Intelligence, Stateless CORE
AGENDA

Introduction
Segment Routing Deep Dive
Segment Routing SDN and Use Case
Summary

Juniper Confidential
Segment Routing Introduction draft-ietf-isis-segment-routing-extensions-xx
Source Based Routing

• Idea from Draft-Kompella( Label Block and Index)

• Network represented by Segment Two Adjs 12

– Adj, Nodal Segment(unique #, one segment)

Kireeti Kompella
– Segments act as topological sub-paths that can be combined 11
1 2 13
together to form the desired path.
– Source Routing: the source chooses a path and encodes it in the 6 7 3 Four Adjs
packet header as an ordered list of segments
• Every Node Forwarding table only take care portion of network 16 5 4 14
– All nodal segment, SRGB(SR Global Block)
– Adj Segment, No neighbors Adj Segment, Local Significant 15
• CSPF for nodal Segment SR Index 11
– Calculate the OIF only,
– label keep same(64-5000 reserved) protocols { isis {
source-packet-routing { node-segment ipv4-index 11}}
Segment Routing Architecture 1
Step1: Build SR Topology by IGP Ext Advertisement 12

1 2 13
11

Segment
Nodes
3
Prefix SID
6 7
Controller

16 5 4 14

15

Adjacency Out
In Label Out intf
Segment Label
100 100 Intf1 Every Node
12 1 2 Node
101 101 Intf1 Share same
Segment Ids
… … …
1 2 6 7 3 111 111 Intf2
11
Adjacency 5001 Pop Intf1 Various
Segment Ids 5002 Pop Intf2 interfaces
6 7 5 4 … … …
5004 Pop Intf2
Segment Routing Architecture 2
Step2: Controller calculate/program Label stacks from Edge

12

Controller
5001 5002
1 2
11
5007 5007

5004
12 5005 66 7 3

5014 5016
1 2 13 5 4 14
PayLoad 11 PayLoad

6 7 3 15

16 5 4 14

15
Adj/Nodal Segment forwarding
Nodal/Adj Label space is different, No Recursive look up.
Packet injected anywhere
1 with label 114 will reach node 14
11
1
11
6 7 3

6 7 3
5001 4 14

5007 5007
4 14
5004 5004 5004

5014 5014 5014 5014 114 114 114

PayLoad PayLoad PayLoad PayLoad PayLoad PayLoad PayLoad PayLoad

• Node Advertise Adj label, IGP extension • Node advertise, unique {64-5000}
• Only install Adj label on router, not aware of rest network. • IGP extension, normal SPF for all loopback
• Push multiple labels stack to reach remote router • Nodal label keep same in every nodes
• POP label only • Swap Label Only
Path Creation 12
Source Based Routing
1 2 13
11
• A. Follow the IGP
– one label pushed, the nodal segment(Node-SID), 6 7 3

– SPF can leverage the ECMP path

16 5 4 14
– Example, {114}

• B. Explicit Via nodal ( like loose node in RSVP-TE) 15

– Push list of via nodal…
12
– Between nodal, SPF load balance.
– Easy to expended across Area/AS 1 2
11 13
– Example, {112,114}
6 7 3
• C. Explicit via Adj, any path
– Push of list of Via Adj 16 5 4 14
– Example, {5001,5002,5003,5004,114}
15
• D. Mixed Path with Adj/Nodal
 ANYCAST SEGMENT ID FOR NODE REDUNDANCY
draft-psarkar-spring-mpls-anycast-segments-01
• Anycast SID
– A group of Nodes share the same SID 5100
Anycast SID: 100
– Work as a “Single” router, single Label 8070 A1, SID: 30 A3, SID: 50 R2, SID: 70
[8000-9000] [8000-9000] [8000-9000]

• Any Topology PAYLOAD 8070

S, SID: 10 R1, SID: 20
– Hub/Spoke [8000-9000] [5000-6000]
D, SID: 80
[8000-9000]
– Ring Topology 5100
8070
– Anycast and other nodes follow IGP

• Application
– ABR Protection A2, SID: 40 A4, SID: 60
R3, SID: 80
[8000-9000]
[8000-9000] [8000-9000]
– Seamless MPLS
– ASBR inter-AS protection
 TI-FRR/TI-LFA
SEGMENT ROUTING CAN GUARANTEE 100%

• IP-based FRR not guaranteed in any topology

• Directed LFA (DLFA) is guaranteed when metrics Backbone Backbone
only cover few cases, extra computation (RLFA) C1 C2 C1 C2
also 90%+ topology
• TI-FRR, Target LDP session with RSVP Tunnel Target LDP
E1 session E4 E1 E4
• TI-LFA Segment Routing, 2 actions
– node segment to P node( From E1, can reach C1 141 99
141 100
without via failure link. E2
E L O O P -F R E E A LT E R N A T E S (R L FA ) E3 E2 E3
– adjacency segment from P to Q Node(From Q 141 X
N D TA R G Enode
T E can
D Lreach
D P C1 S Ewithout
S S IO via
N Sfailure Link)
– TI-LFA
Node protection for remote LFA vs. local LFA only
100% Guarantee Node SID to P node,
100
90
Follow Adj SID to Q
80
70
60
50
40 IP FRR Segment Routing FRR
30
20
10
0
LFA R-LFA TI-LFA
Remote LFA Local LFA

600
incoming T-LDP sessions
Binding SID in Multi-Area SR, Larger network w/ Label stacks
Advertising LSPs from other protocols into SPRING
Global node label = 120
RSVP LSP to reach R31 with
ERO=R33,R34,R31 (use local label 500)
RSVP LSP
120
500
R20 R33
Pay R30
Load
R21
Global node label = 200
SR-LSP to reach R30 with
ERO=R32,R31,R30 (use local label
510)
R34 R31

200
510 R32
332
Pay SR LSP
Load 331
330
Pay
Load

RSVP
SPRING SPRING
SRV6 STANDARDIZATION

• IETF is in the process of standardizing SRv6

– Draft-ietf-6man-segment-routing-header-01
– Work in Progress

• Two modes of operation

– Insertion mode
 SR ingress router inserts an SRH between IPv6 header and IPv6 payload
 SR egress router optionally removes the SRH
– Prepending mode
 SR ingress router prepends a new IPv6 header and an SRH to the original IPv6 header
 SR egress router always removes the new IPv6 header and the SRH, leaving only the original IPv6 header
Segment Routing IPv6(Animated)
include a SRH, Insertion mode and Prepending mode
Ver DSCP Flow Label IPv6
Source
2001:db8:0:1::1 Length Next HDR Hop Limit HEADER
136
136
80 SRH
TCP
SRH 252
251
254
253
255
250
249
SRv6 Ingress Source Address
2001:db8:0:1::2 2001:db8:0:1::1
SRv6 Router Destination Address
2001:db8:0:1::3 2001:db8:0:1::6
2001:db8:0:1::4
2001:db8:0:1::4
2001:db8:0:1::3
2001:db8:0:1::5
IPv6
Router
Next HDR
TCP
Length
56
HDR Type
4
Seg Left
2
13
TCP
Segment
Header
First Seg Flags Routing Header
2 Reserved
IPv6 C=1
Router Segment 0
2001:db8:0:1::6
SRv6 Router
SRv6 Egress 2001:db8:0:1::4 Segment 1
2001:db8:0:1::5
2001:db8:0:1::5
Segment 2
Destination
2001:db8:0:1::6 2001:db8:0:1::4

TCP Header
Draft-ietf-6man-segment-routing-header-01
 SEGMENT ROUTING SDN
WORK GREAT WITH SDN &PCEP
1
11

Tunnel onto
{11, 1, 3, 14} 6 7 3
Path 11-1-7-3-14 is ok.
I account the BW.
Then I steer the traffic 14
Controller
101 4
on this path
103 103
Segment Routing 114 114 114
SDN Controller
PayLoad PayLoad PayLoad

• The network is simple, highly programmable and responsive to rapid changes

• Source Based routing, label pushed in the source will decide the path.
• On router, PCE Client no need signaling protocol to create path, Just Segment Routing.
• Better than PCE+RSVP-TE, No on-demand signaling the path.
• Better than Static MPLS label push from SDN, SR still have ECMP, Resilience, FRR.
Segment Routing vs LDP/RSVP
Keep the network Status Simple, Build the network topology

Segment Routing LDP • Segment Routing

– Only keep minimal status in network
12 12
– Keep all loopbacks
– With only adj prefix
1 2 11 1 2 13
11 13 – One SPF for all nodal ID.
6 7 3
6 7 3 • LDP
5 4 14 16 5 4 14 – Keep all Loopbacks
16
– Adj Prefix and non-adj prefix
15 15
• RSVP
– Keep all Loopbacks
Non Adj Prefix
– Potential full mesh LSP, and middle node
Adj Prefix Adj Prefix
keeps a lot of transit information per LSP.
Loopback Loopback – Per LSP CSPF caculation
– Known as not so scale protocol.
Segment Routing vs MPLS

Features MPLS Segment Routing

LDP/RSVP/BGP( any of label allocation) OSPF/ISIS,
Control Protocol OSPF or ISIS or BGP, or SDN Controller
BGP ( any of topology), SDN

Traffic Engineer RSVP, PCE Client, SDN OSPF/ISIS(option) SDN (option)

Fast Reroute LDP FRR, or RSVP-TE FRR Build in FRR, cover for all scenario

With help of BGP label, or RSVP-TE inter Area

Inter-Area/Inter-AS Loose Node ID extension
hard to protect

Source Path Routing No, IGP only Yes, explicit indicate ingress

Scalabilities LDP same as IGP….RSVP limited. Node + ADJ Segment(less entry) Best Scale

Performance
NO Build in with RFC 6374
Measurement

SDN integration PCE, RSVP-TE PCE, BGP-LU, SR

SEGMENT ROUTING FOR CLOUD DEPLOYMENT
UNDERLAY PATH BY SR PROTOCOL, OVERLAY SDN CONTROLLER WITH LABEL APP

Monitoring & App Build & Pkg Overlay Virtualized

Analytics

LB FW
Virtual Virtual
Network Network

App Test & Deploy Network Services APP

DevOps
Segment List
Containers
BMS
vRouter
label for
App/Dockers
or VPN etc.
Orchestrator /
Controller / Tools Segment
List for
Monitoring & Path
Troubleshooting

Underlay Set-up

Automated
Provisioning Ops

Underlay Physical
AGENDA

Introduction
Segment Routing Deep Dive
Segment Routing SDN and Use Case
Summary

Juniper Confidential
SPRING : DOMAIN APPLICABILITY

Fixed design, EBGP as IGP,

Data Center Simpler mgmt. with common Controller
SRGB
Label stack topology
Alternate way of doing FRR,
WAN No core state, BGP-LS to
export topology to controller

FRR in Metro rings, PW

Metro transport
Ingress Node Egress Node
Traffic engineering,
WAN/Metro Core
Edge Northbound interface: PCEP,
BGP-LU, Flow-spec
PCE WITH SEGMENT ROUTING
PCE-initiated LSP :
draft-ietf-pce-segment-routing-07

• PCEP SR similar with RSVP-TE PCEP 1

Service
Request
– Open message negotiate SR-PCE-CAPABILITY TLV
– PCCreate LSP with SR-ERO for Label stack
– No Need Signaling on PE-P-PE
PCEP Controller
– LSP State report with SR-RRO

• BGP-LS get the network information 2 PCCreate LSP 5 Delegate

– TEDB information with label send back to Controller With SR-ERO
4 LSP State report

BGP
– draft-gredler-idr-bgp-ls-segment-routing-ext-xx.txt w/ SR-RRO
PCC
3 No Signaling
• Service mapping by
PBR, QPPB
– Openflow/PBR/QPPB/BGP FlowSpec BGP Flow Spec
PE1 ASBR
Open Flow
BGP FlowSpec redirect to SR LSP Tunnel
Type Matching Type Matching
Type 1 Destination prefix Type 7 ICMP type
Type 2 Source prefix Type 8 ICMP code
Type 3 IP protocol Type 9 TCP flag
Type 4 Port (Defines a list of pairs that matches source or Type 10 Packet length
destination UDP/TCP ports)
Type 5 Destination port Type 11 DSCP
Type 6 Source port Type 12 Fragment

Type Extended Community Encoding

0x8006 Traffic-rate 2 byte/4 byte float
0x8007 Traffic-Action bitmask
0x8008 Redirection 6-bye route-target
0x8009 Traffic-marking DSCP Value

NOTE: Detailed information about each type and filed can be found in RFC 5575 section#4 “Dissemination of Information”.
Segment Routing with PCEP and BGP-LS
 Prefix & node SID learning via ISIS &/or BGP-LS
 New PCEP capability, ERO subobject and TLVs
 draft-ietf-pce-segment-routing-06
 SPRING-TE LSP creation, visualization & optimization
 draft-rosen-idr-rfc3107bis-00.txt
BGP-LU WITH SEGMENT ROUTING NOT draft-ietf-idr-bgp-prefix-sid-03

Service
1
Request

• BGP-LU Session between Controller/Router

– BGP LU carrier the label stack for SR/LSP
2 BGP-LU with
– BGP-LU carrier the Label stack for LSP + VPN Service BGP-LU Controller
Label Stacks
101

• BGP-LS get the network information 103

114
– TEDB information with label send back to Controller

BGP
80001
– draft-gredler-idr-bgp-ls-segment-routing-ext-xx.txt BGP-LU
PayLoad

• BGP is the only protocol for Service and

Tunnel PE1 ASBR

– QPPB/BGP FlowSpec Example from ExaBGP

bespalov@CentOS-1 ~/exabgp-3.4.16/sbin>cat ~bespalov/config/exabgp neighbor 192.168.255.12 {
local-address 192.168.255.2;
– With additional Openflow/PBR peer-as 65000; local-as 65000;
family { ipv4 nlri-mpls; }
static {
route 10.255.255.8/32 {
next-hop 10.0.0.2;
label [ 800005 800007 800006 800008 ]; }}
MPLS IN DATA CENTERS
• Overlays are widely used today
– South → North: Egress Peer Engineering (EPE)

– North → South: Load balancing, Floating IPs, ...

– East ↔ West: Multi Tenancy

• Currently overlays are IP-based, moving to MPLS

– Consistent end-to-end protocol; avoid ‘impedance-mismatch’ at boundaries

– Hierarchical Forwarding [MPLS Label Stack]; reduces FIB state

• Use SPRING-like approach

– Label stacking (hierarchy) to reduce FIB size on switches with merchant silicon

– Label stacking for ‘source-routing’ across WAN

– Different control plane inside data-center: BGP instead of IGP

SPRING INTRA DATA CENTER ROUTING 1

Controller

Proprietary BGP-LU + SPRING BGP-LU + SPRING Proprietary

Egress VM + Egress TOR Label Egress Server Label Egress VM Label
Egress Server + Egress TOR Prefix-SID Egress Server Prefix-SID
Egress TOR SRGB SRGB
Label stack

VM VM

MPLS label MPLS label MPLS label

Payload
Egress VM Egress server Egress TOR
SPRING INTRA DATA CENTER ROUTING 1

MPLS label MPLS label MPLS label

Payload
Egress VM Egress server Egress TOR

"Loose route"
ECMP ECMP over spine switches
Egress TOR

VM Egress VM
Egress server
BGP-LU PREFIX SEGMENT PROPOSAL
Juniper Proposal [draft-gredler-idr-bgplu-prefix-sid-00]

Controller
BGP-LU
BGP-LU BGP-LU BGP-LU FEC: F
X X
FEC: E FEC: E FEC: E Label: 1001
FEC: G FEC: G
Label: 4005 Label: 1005 Label: null
Label stack: Label: 300
300, 1001, SRGB:4000- SRGB:1000- SRGB:6000-
4005(top) 5000 2000 7000
SID: 5 SID: 5 SID: 5

VM VM

300, 1001, 300, 1001, 300, 300(top)

A B 4005(top) C 1005(top) D 1001(top) E F G
 SPRING INTER DATA CENTER ROUTING 1

MPLS label MPLS label MPLS label MPLS label stack MPLS label
Payload
DC2 Egress VM DC2 Egress server DC2 Egress TOR DCI path: A, B, C, D DC1 Egress Router

DC1 DC2

ECMP B DC2
Egress
TOR

DC1 A DCI C D
Egress
Router DC2
= Anycast Group Egress VM
VM
DC2
Egress
Server
SPRING INTER-DOMAIN CLOUD TRAFFIC ENGINEER 2

1 Cloud Traffic engineer

CDN

BRANCH HOME

MOBILE HQ

2 Fish Topology
SP DC CDN

Easy to optimize End-To-End Traffic for SP Owned Network.

How to optimize VIP Customer for Internet/Cloud connection?
BGP EPE DESIGN PHILOSOPHY 2
How to Select Which Peer to send
 Controller/RR may morning the BGP Peer Link
Peer
 Controller/RR find a tunnel from Ingress to ASBR
Controller
 Controller/RR based on certain rules to select ASBR
Peer
How ASBR identify a Peer
 Per Peer /32 address per label LDP Peer
 Install the MPLS Label POP for every Peer
Segment
 When ASBR received different label and Routing LU 100
send traffic to specific Peer Peer
GRE
How Ingress mapping traffic to ASBR/Peer MPLS IP Forwarding

 Ingress push tunnel label to ASBR

BGP-LU w/ Label 100 BGP
 Ingress push BGP-LU label MPLS Label
Push
LU Label LU Label
Push
100 100 POP
Payload Payload Payload
BGP-LU EPE & MPLS KEY BENEFITS
EXTEND HOLLOW CORE/LSR TO PEERING, CHEAPER PEERING SOLUTION 2

Netconf/Yang BGP LU/LS

MP-BGP EVPN Controller Segment Routing
NO IP Lookup!
Can be
Normal IP forwarding!
<128K Prefixes

BRANCH HOME
LSR
Core

TOR/BNG/PE
MOBILE HQ

MPLS LER MPLS LSR IP Forwarding

Push SR Label SWAP SR Label POP

Push LU Label LU Label LU Label POP

Payload Payload Payload Payload Payload

SEGMENT ROUTING AND EPE USE CASE 2
Select Peer Select Egress Router

Customer packet Meta-data Tunnel encapsulation

Peer

Controller
Peer
TOR Leaf Spine

VM Peer
Server
Content Provider
Data Center
WAN
Content Provider Data Center

Customer packet Meta-data Tunnel encapsulation

Floating IP for service Select VM Select Server
SEGMENT ROUTING IN ACCESS/AGGREGATION
3
SIMPLIFIED BOX FUNCTION, MOVE INTELLIGENCE TO CONTROLLER

Controller

EVPN
BGP FlowSpec BGP-TE/SR BGP- VPWS IPv4
EVPN 6vPE IPv4 VPN

Service
Openflow 3107 VPLS IPv6

BGP

Transport
SR/IS-IS
BGP-LU IS-IS OSPF Static RSVP LDP
MPLS
Forwarding MPLS Forwarding

• Keep OAM/Clocking
• No need Peer with others, only Controller
• No Need Compute, Controller got full network view.

Minimal Protocols, Dumb Box in Access

 SEAMLESS MPLS EVOLUTION – SEGMENT ROUTING
3

• Architect Change
– To manage 1,000+ boxes Add SDN Controller

Controller NETCONF BG—LU

for VPN Service for Tunnel
– RSVP-TE w/ RFC3107 to Segment
BGP-LS
Routing for Infor
Mobile Terminals
• Technical Benefits Access Metro DC
– SP Fabric management with ZTP
– Better FRR with LFA/RLFA/TI-LFA leaf Agg
– Better ABR Node protection with
Home
Segment Routing Anycast SID or SOHO
leaf
– Better tunnel provision by BGP-LU or Service Edge
Controller Router
Branch
– Better Tunnel Stitching by SR, no need Office
RFC3107, save one label
leaf
– Service Provision by NETCONF Service Edge vBNG vEPC
– Network information collect by BGP-LS Agg
Router
HQ leaf
SEGMENT ROUTING FOR NFV SERVICE CHAINING
4
NO NEED NETWORK SERVICE HEADER(NSH), VNF SUPPORT MPLS
Shortest path
Tunnel
Forwarding path

Services provided off-path by physical or virtual service nodes

Packets diverted through tunnels
 Return to forwarding path
 By tunnel
 Via forwarding
 After attention by other service nodes
SEGMENT ROUTING FOR NFV SERVICE CHAINING
4
NO NEED NETWORK SERVICE HEADER(NSH), VNF SUPPORT MPLS
S2
Shortest path Y
Tunnel
Forwarding path Pay
Load Y
Pay
Load

Pay
Load

X
S1 X S1
S2 S2
Y Y
Pay Pay
Load Load
Push label Stack for Service Chaining.
VNF support MPLS label
 TELCO CLOUD
WHAT IS THE TELCO CLOUD ARCHITECTURE? HIGH LEVEL ARCHITECTURE 5
Connectivity Telco Cloud
Key Properties
Building Blocks
WAN/METRO
1. Physical distribution providing BGP (Control Plane) WAN

fungible cloud resources close to MPLS (Service)

MPLS (Transport)
Telco consumer and business
eyeballs.
+
DC Fabric
WAN

BGP / OSPF (Control Plane)

IP (Transport) METRO
2. Enables applications to have: WAN
1. Low Latency +
DC Overlays WAN
2. High Availability (through MPLS, VXLAN, IP, GRE, etc.
WAN
METRO
distribution) WAN
3. High volume of last mile METRO
throughput; minimizing network BGP
MPLS,
VXLAN
Service WAN
wide capacity growth (choke points) L3VPN,
BGP
Overlays

EVPN WAN
~50-250
3. Seamless Integration of DC and IP or MPLS
Transports ~250-1000
WAN technologies leveraging
> 1000
existing network and operational
procedures.
Openstack
Neutron & Neutron Extensions,
etc.
TELCO CLOUD HIGH LEVEL REQUIREMENTS 10K FEET
5
MPLS in SP Fabrics - High level vision
Fabric Fabric
A P
L
BGP-SR L P
BGP-SR A
Metro Core Metro L L
C 3107 3107
L
P P P P P P
Fabric L OSPF-SR L L RSVP-TE L L OSPF-SR L Fabric
A P
L
BGP-SR L P
BGP-SR A
L L
C D2F
L

Fabric Fabric
A P
L
BGP-SR L P
BGP-SR A
L L
C
L

EVPN SR
• Underlay transport is based on Segment Routing
• EVPN Signaling is a key requirement for *all* control • No IGP in Telco Cloud. Only BGP-LU with prefix-SID
plane signaling extensions
• EVPN-VPWS with flexible-cross-connect for all L2 • Metro moves to OSPF-SR
pseudowires
• EVPN-MPLS multi-point with IRB
• EVPN-VXLAN for for IP fabrics
 JET w/
6 SPRING
STATIC SEGMENT ROUTING
Step1: Build the Segment Routing Topology, Single Hop LSP

Adj_sid_23: Adj_sid_34:
Adj_sid_45:
in_label =1000001 in_label =1000002
in_label =1000003
Pop Pop
Pop
Nexthop = R3 Nexthop = R4
Nexthop = R5

CE1 R5 CE2
R1 R2 R3 R4
Adj_sid_21: Adj_sid_32: Lsp_41:
In_label = In_label = In_label = 10100001
10100003 10100002 Swap
Pop Pop Out_labels = 10100003, 10100002
Nexthop = R1 Nexthop = R2 Nexthop = R3
 JET w/
6 SPRING
STATIC SEGMENT ROUTING
Step2: Push the SR LSP from Edge

Lsp_15: Lsp_51:
Dest = R5 Dest = R1
Push Push
Out_labels = 10000003, 10000002, 10000001 Out_label =
Nexthop = R2 10100001
Nexthop = R4

Ingress LSP with a stack of Adj-SID labels:

destJnxP = IpAddressAddrFormat("128.9.148.133")
dest = JnxBaseIpAddress(destJnxP)
CE1 R5 CE2 lsp = RoutingStaticLspEntry()
R1 R2 R3 R4
Adj_sid_45: lsp.name = “lsp_15”
Adj_sid_23: Adj_sid_34:
in_label in_label =1000002 in_label =1000003 lsp.type = 0 << ingress
=1000001 Pop Pop lsp.Prefix = StaticLspEntryPrefix()
Pop Nexthop = R4 Nexthop = R5 lsp.Prefix.destination = dest
Nexthop = R3 lsp.label_operation = 0 << push
Adj_sid_32: Lsp_41:
In_label = 10100001
lsp.outgoing_labels = ["1000003","1000002","1000001"]
Adj_sid_21: In_label =
Swap lsp.nexthop = "55.1.12.2"
In_label = 10100002
10100003 Pop Out_labels = 10100003, 10100002 lsp.preference = "6"
Pop Nexthop = R2 Nexthop = R3 lsp.metric = "1"
Nexthop = R1 addReq = RoutingStaticLspAddRequest(lsp)
addReply = staticLsp.StaticLspAdd(addReq)
print 'Reply status = ', addReply.status
OPENFLOW WITH SEGMENT ROUTING
ONF's SPRING-OPEN

• OpenFlow 1.3.4 can push 2 labels

– Service label and Tunnel labels
– Use Openflow group Chain to push multiple labels

• Openflow Build the Segment Routing Topo

– Adj SID for POP
– Node SID for continue(no change/no swap)

• No RSVP-TE/LDP and IGP on those routers

– Only MPLS dataplane and Static configure from Openflow

• A lot of limitations BUT can show

– Intelligence on Controller, very ugly CLI on Controller
– White Label box with simple MPLS forwarding Plane
– Demo in Dec 2014. https://goo.gl/ddeX5N
AGENDA

Introduction
Segment Routing Deep Dive
Segment Routing SDN and Use Case
Summary

Juniper Confidential
Summary- Segment Routing Re-Invent MPLS

• Seamless work with SDN, BGP-LU/PCE-P Architecture.

instantly tunnel setup. for next generation Application SDN
driven networks
• Work with NFV, such as Service Chaining
Service BGP-LU
• Simplified MPLS Control Plane, OSPF/ISIS only. No need
Signaling for tunnel setup. Tunnel path decided by Chaining PCEP
ingress router. Segment
– source routing and hence explicit routing Routing
• less status inside network
element(router/switch)Topology based on Adj/Nodal
information. Independent with Application Status MPLS
Forwarding
100% FRR
• 100% IP fast reroute protection, Fit for any topology
• Work great with Traffic Engineer and IPv6.. With QoS,
OAM/SLA
Segment Routing Customers
Re-invent MPLS again! Foundation of NFV/SDN
• Major vendors claim to support, ALU/Cisco/Huawei/Juniper
• Known customer transforming to SPRING
• AT&T CORD
• Microsoft SWAN
SDN NFV
• China OTT, Tencent/Alibaba
• Japan Softbank/NTT
• ANZ Telstra etc Underlay

CPE Access Edge Core DC

ROAD TO SELF DRIVEN NETWORK

SDN/NFV
Controller

Network
Segment Routing
Telemetry
Network
SUMMARY

1 Segment Routing Design for SDN

2 Segment Routing simplify Protocols

3 Segment Routing enable better traffic engineer, IGP/BGP, Egress Peering Engineering

4 Segment Routing Provide better FRR protection

5 Segment Routing can be deployed in All Domains, DC, Metro, Access, Telco Cloud etc.
THANK YOU

JUNIPER CONFIDENTIAL