Transcript Packets & Photons: The Emerging Two Layer Network

Packets & Photons:
The Emerging Two
Layer Network
October 2001
Copyright © 2000, Juniper Networks, Inc.
1
Agenda
 History
 The
of IP Backbones
Emerging Two Layer Network
 Network
Platforms
 Standards
and Forums
 GMPLS
2
Typical IP Backbone (Late 1990’s)
Core
Router
Core
Router
ATM
Switch
ATM
Switch
MUX
SONET/SDH
ADM
MUX
SONET/SDH
ADM
SONET/SDH
DCS
SONET/SDH
DCS
SONET/SDH
ADM
MUX
MUX
ATM
Switch
ATM
Switch
Core
Router

SONET/SDH
ADM
Core
Router
Data piggybacked over traditional voice/TDM transport
3
Why So Many Layers?

Router

MUX
Packet switching
 Speed match router/ switch
interfaces to transmission
 Multiplexing and statistical
network
gain
 SONET/SDH
 Any-to-any connections
 Restoration (several seconds)
 Time division multiplexing
(TDM)
 ATM/Frame switches
 Fault isolation
 Hardware forwarding
 Restoration (50mSeconds)
 Traffic engineering



Restoration (sub-second)
DWDM
Raw bandwidth
 Defer new construction


Result
 More vendor integration
 Multiple NM Systems
 Increased capital and operational costs
4
IP Backbone Evolution
Core
Router
(IP/MPLS) 
FR/ATM
Switch
MUX
SONET/SDH
MUX becomes redundant
Core
Router
(IP/MPLS)
 IP trunk requirements
reach SDH aggregate
levels
 Next generation
routers include high
speed SONET/SDH
interfaces
SONET/
SDH
DWDM
DWDM
(Maybe)
5
IP Backbone Evolution
Core
Router
(IP/MPLS)
 Removal
of ATM Layer
 Next generation routers
FR/ATM
Switch
MUX
provide trunk speeds
 Multi-protocol Label
Switching (MPLS) on
routers provides traffic
engineering
Core
Router
(IP/MPLS)
SONET/
SDH
SONET/SDH
DWDM
DWDM
(Maybe)
6
Removing the ATM Layer
Logical
Topology

Why Remove ATM?
 Two networks to manage - IP and ATM
 Cell tax
 Lack of high-speed SAR interfaces
 High density of virtual circuits
 IP routing protocol stress
7
Agenda
 History
 The
of IP Backbones
Emerging Two Layer Network
 Network
Platforms
 Standards
and Forums
 GMPLS
8
Collapsing Into Two Layers
IP Service (Routers)
Optical Core
Optical Transport
(OXCs, WDMs,
SONET ?)
9
Collapsing Into Two Layers
IP Service (Routers)
Optical Core
Optical Transport
(OXCs, WDMs,
SONET ?)

IP router layer functions








Service creation
Multiplexing and statistical gain
Any-to-any connections
Traffic engineering
Restoration (10s ms)
Subscriber to transport speed matching
Delay bandwidth buffering and congestion control
10
Internet scalability
Collapsing Into Two Layers
IP Service (Routers)
Optical Core
Optical Transport
(OXCs, WDMs,
SONET ?)

Optical transport layer functions
 TDM and standard framing format
 Fault isolation and sectioning
 Restoration (10’s ms)
 Survivability
 Cost efficient transport of massive bandwidth (DWDM)
 Long haul transmission distances
 Metro transmission distances ????
11
The Emerging Two-Layer
Network
Data Layer
Routers
IP Services
Transport Layer
OXC’s
TDM’s
WDM’s
LH Transport
Reduced cost
 Transport layer visible to IP
Services
 Transport layer signaling is
an open standard (RSVP &
CR-LDP)

Reduced complexity
 Network more scalable
 Uniform admin &
management of IP and
transport layers

12
Agenda
 History
 The
of IP Backbones
Emerging Two Layer Network
 Network
Platforms
 Standards
and Forums
 GMPLS
13
SONET/SDH Benefits

Rapid and predictable restoration
 10s of ms; depends on ring size
 Simple to engineer
Standard framing and multiplexing
(Time Division Multiplexing [TDM])
 Maintainability

 Performance monitoring
 Fault isolation and sectioning
 Bandwidth management
 Network management

Transparency
 Voice, video or data traffic

Challenge
 Remove complexity
and keep benefits
14
SONET/SDH Benefits

Rapid and predictable restoration
 10s of ms; depends on ring size
 Simple to engineer
Standard framing and multiplexing
(Time Division Multiplexing [TDM])
 Maintainability

 Performance monitoring
 Fault isolation and sectioning
 Bandwidth management
 Network management

Transparency
Traffic
Quickly
Rerouted
After Failure
 Voice, video or data traffic

Challenge
 Remove complexity
and keep benefits
15
SONET/SDH Limitations

Traditional SONET/SDH-based networks
 Engineered for voice, not data
 Slow to provision
Planning complexity
 Grooming complexity
 Delivery measured in weeks

 Expensive to scale

Space, power, one wavelength per chassis
 Inflexible


Static not dynamic bandwidth
Granularity – why not 5.5Gbps ?
 Little interoperability at “control plane”
Customers forced to buy from one vendor
 Stifles “best-in-class” deployment


Packet layer – no visibility into optical layer
16
What is an IP Router?
A Device Which Moves IP Datagrams Across
an Internetwork From Source to Destination
 Minimum qualifications
ISO 7 Layer Model
 Capable of switching IP datagrams:
L3 forwarding
 Symmetric any-port-to-any-port
switching speed
 Delay-bandwidth buffering,
plus congestion control
 Internet scale IS-IS, OSPF, MPLS, BGP4
7 - Application
6 - Presentation
5 - Session
4 - Transport

Today’s benchmark
3 - Network
 Wire-rate forwarding on all ports
2 - Datalink


1 - Physical


for 40-byte packets
Performance independent of load
Support of CoS queuing, shaping,
and policing
Traffic engineering
Classification and filtering at wire rate
17
What is an IP Router?
Routing Algorithm Goals
 Optimal routes
 Calculate and select the best routes –
many methods
ISO 7 Layer Model
7 - Application

 Functional efficiency with low routing
protocol overhead
6 - Presentation
5 - Session

in a variable environment (hardware
failure, high load, topology changes)

Rapid convergence
 Slow route calculations cause loops
2 - Datalink
1 - Physical
Robust and stable
 Predictable and correct functionality
4 - Transport
3 - Network
Simplicity
and drops in service

Flexibility
 Speed + accuracy to adapt to network
changes (bandwidth, delays, queues,
traffic levels, etc.)
18
What is an IP Router?
IP Service Creation

 Internet scale routing allows anyone
ISO 7 Layer Model
to connect to anyone
(within or outside of own company)
7 - Application
6 - Presentation

HTML from FTP

Multicast
 Not possible with voice circuit switching
technology
 Internet radio, video on demand,
push Web
3 - Network
2 - Datalink
1 - Physical
Applications
 Processing granularity to differentiate
5 - Session
4 - Transport
Any-to-any connectivity

Content sites
 Directing Web traffic
 Complementing cache servers
 Security
19
Optical Cross-connects (OEO)
SONET/SDH
Digital Cross-connect (DXC)
Also known as Digital
Cross-connect Switch (DCS)
DXC/DCS
20
Optical Cross-connects (OEO)
SONET/SDH
Digital Cross-connect (DXC)
Also known as Digital
Cross-connect Switch (DCS)
ATM
Electrical
Switch
Matrix
STS-N
DS-3
STS-N
ATM
STS-N
DS-1
ATM
DS-3
DS-3
DS-1
DS-1
DS-1
STS-1
STS-N
STS-1
STS-N ATM DS-1 DS-3
21
All Optical Cross-connects
(OOO)
All Optical Cross-connect (OXC)
Also known as Photonic
Cross-connect (PXC)
OXC/PXC
22
All Optical Cross-connects
(OOO)
l2
l4
All Optical Cross-connect (OXC)
Also known as Photonic
Cross-connect (PXC)
l1
l3
Optical
Switch
Fabric
l3
l4
l1
l2
23
What is an Optical
Cross-connect?
Connects one port (l) to another
port
 Add/Drop function with certain l
 Delivers high bandwidth
 Quick to provision bandwidth

ISO 7 Layer Model
7 - Application
6 - Presentation
5 - Session
4 - Transport
3 - Network
l1
Port 1
Port 3
l2
2 - Datalink
1 - Physical
Port 2
Port 4
l2
l1
24
OXC/PXC Switching
Mechanisms

Fibers

Micro-electrical Mechanical Systems
MEMs
 Used for many other applications
Reflector


Imaging
Lenses

From Lucent, Corning, Xros (Nortel),
and others
Currently 8 x 8 OXC
256 mirrors, long-term goal 1,024
 OXC
 ADM uses seesaw MEMS

Electrical controls
 Voltage applied to mirror; tilts on 2
MEMs tilting mirrors
axis + or – 6 degrees

Switch times typically 10 to 25 ms
25
OXC/PXC Switching
Mechanisms
 Liquid
Crystal Light Valves
 From Spectra
Liquid Crystal Cell
Switch and Chorum
ON
technologies
Output 1
Input
 Switch speed
sub-millisecond
 Future switch
speed in nanosecond
 1 x 2 port switch
Polarizing
Polarizing
Beam
Beam
 2 x 2 Add/Drop
Splitter
Splitter
 Electrical controls
Liquid Crystal Cell
26
OXC/PXC Switching
Mechanisms
 Liquid
Crystal Light Valves
 From Spectra
Liquid Crystal Cell
Switch and Chorum
technologies
Input
 Switch speed
sub-millisecond
 Future switch
speed in nanosecond
 1 x 2 port switch
Polarizing
Polarizing
Beam
Beam Output 2
 2 x 2 Add/Drop
Splitter
Splitter
 Electrical controls
OFF
Liquid Crystal Cell
27
OXC/PXC Switching
Mechanisms
Bubbles



From Agilent
32 x 32 or dual 16 x 32 ports
Suitable for
 Wavelength Interchange
Cross-connect (WIXC)
 Wavelength Selective
Cross-connect (WSXC)
 Optical Add/Drop Multiplexers
(OADM)

Inkjet + Silica Planar
Lightwave Circuitry
 Electrical controls
 Bubbles created by heating
“index matching fluid”

Switch times under 10 ms
28
Developing an All Optical
Packet Router

Needs
 How do you read a photonic header?
 The
“pipeline” approach?
 Switching and logic
 Current
technology not fast enough
 Lithium Niobate devices have speed,
but too much crosstalk
 Photonic Bandgap Devices
(optical equivalent to transistor)
 Buffering/Memory
 Optical
buffers (fixed loop delays) exist,
but are insufficient
 Bi-stable lasers
 Holographic memories
 SEEDS (Self Electro-optic Effect Devices)
29
Agenda
 History
 The
of IP Backbones
Emerging Two Layer Network
 Network
Platforms
 Standards
and Forums
 GMPLS
30
Operational Approaches:
Overlay and Peer Models

Overlay model
 Two independent control planes

IP/MPLS routing

Optical domain routing
 Router is client of optical domain
 Optical topology invisible to routers
 Routing protocol stress – scaling issues
 Does this look familiar?

Peer model
 Single integrated control plane
 Router and optical switches are peers
 Optical topology is visible to routers
 Similar to IP/MPLS model
31
?
Operational Approaches:
The Hybrid Model

Hybrid model
 Combines peer & Overlay
 Middle
ground of 2 extremes
 Benefits
of both models
 Multi admin domain support
 Derived
from overlay model
 Multiple technologies within
domain
 Derived
from peer model
Peer
UNI
32
Standards and Industry
Forums

Optical Internetworking Forum (OIF)
 Industry forum
 Kick-off meeting May 1998
 Standard OIF UNI based
on IETF work (CR-LDP/RSVP)

Internet Engineering Task Force (IETF)
 Driving GMPLS standards development
 Initial application was MPlambdaS
 Peer model and Hybrid model
 Extend MPLS traffic engineering
to the optical control plane
Rapid provisioning
 Efficient restoration


ITU-T
 Study Group 13
 Study Group 15
33
IETF

GMPLS now Hosted by CCAMP WG
 Common Control And Measurement Plane


MPLS WG revised charter (without GMPLS)
Eleven main GMPLS building blocks
 Internet Drafts


Current work includes extending existing control
protocols (example, OSPF & ISIS)
New & future extensions considered
 BGP4
 For
cross AS, and Carrier of Carriers applications
 LCAS
 Link
Capacity Adjustment Scheme protocol for SONET
 SONET Virtual Concatenation (dynamic TDM circuit control)

Intent to submit work to ITU-T
34
ITU-T
 Study
Group 13 (SG13)
 Focus: Multi-protocol & IP-based networks &
their inter-working
 Study
Group 15 (SG15)
 Focus: Optical & other transport networks
 G.ASON – Automatically Switched Optical
Network
 Addresses
networks
 Ambition
the control layer for intelligent optical
to reference IETF standards
35
OIF Optical UNI Signaling
OIF-UNI
UNI
UNI
IETF-GMPLS
UNI
UNI
Optical
Transmission
UNI
UNI
Network
Uses procedures and messages defined for MPLS traffic
engineering and GMPLS
 Features

 Runs in UNI-only mode (overlay model)
 Optical path creation, modification, and deletion
 Optical path status inquiry and response

Allows one protocol to support two different applications
 OIF UNI: client bandwidth requests (hide optical topology)
 GMPLS: service provider provisioning (expose optical topology)
36
Agenda
 History
 The
of IP Backbones
Emerging Two Layer Network
 Network
Platforms
 Standards
and Forums
 GMPLS
37
Traditional MPLS Applications
Traffic Engineering
Source
Destination
Layer 3 Routing
VPNs
CPE
FT/VRF
Traffic Engineered LSP
PE
PE
P
Site 1
FT/VRF
CPE
Site 3
FT/VRF
CPE
CPE
P
Site 2
P
CPE
Site 2
P
CPE
FT/VRF
Site 1
Site 3
FT/VRF
PE
P
38
PE
FT/VRS
FT/VRF
Generalized MPLS (GMPLS)


Traditional MPLS supports packet & cell switching
Extends MPLS to support multiple switching types
 TDM switching (SDH/SONET)
 Wavelength switching (Lambda)
 Physical port switching (Fiber)




Peer model
Uses existing and evolving technology
Facilitates parallel evolution in the IP and optical
transmission domains
Enhances service provider revenues
 New service creation
 Faster provisioning
 Operational efficiencies
39
GMPLS Mechanisms
IGP extensions
 Forwarding adjacency
 LSP hierarchy
 Constraint-based routing
 Signaling extensions
 Link Management Protocol (LMP)
 Link bundling

40
IGP Extensions



OSPF and IS-IS extensions
Flood topology information among IP routers and OXCs
New link types
 Normal link (packet)
 Non-packet link (TDM,
l, or fiber)
 Forwarding adjacency (FA-LSP)
41
IGP Extensions



OSPF and IS-IS extensions
Flood topology information among IP routers and OXCs
New link types
 Normal link (packet)
 Non-packet link (TDM, l, or fiber)
 Forwarding adjacency (FA-LSP)
42
IGP Extensions
New Link Type sub-TLVs
 Link protection
1:1 Protection
 Protection capability
Working
 Attributes
 None,
1+1, 1:N, or ring
 Priority for a working
channel
Protection
1:3 Protection
Working
Protection
43
IGP Extensions
New Link Type sub-TLVs
 Link descriptor
1:1 Protection
 Characteristics of the link
 Selected attributes
 Link type
 SONET, SDH, clear,
Gig E, 10 Gig E
Working
Protection
 Minimum
reservable bandwidth
 Maximum reservable bandwidth 1:3 Protection
 Attributes change over time
Working
 Provides a new constraint
for LSP calculation

Shared Risk Link Group
(SRLG)
 List of the link’s SRLGs
 Does not change over time
44
Protection
Forwarding Adjacency
Ingress Node
(Low Order LSP)
Egress Node
(Low Order LSP)
SONET/SDH
ADM
ATM
Switch
A
FA-LSP
Ingress Node
(High Order LSP)
SONET/SDH
ADM
Egress Node
(High Order LSP)
node can advertise an LSP into the IGP
 Establishes LSP using RSVP/CR-LDP signaling
 IGP floods FA-LSP
 Link state database maintains conventional links and FA-LSPs
A second node wanting to create an LSP can use an FA-LSP as a”link”
in the path for a new, lower order LSP
 The second node uses RSVP/CR-LDP to establish label bindings for the
lower order LSP

45
ATM
Switch
Forwarding Adjacency
Ingress Node
(Low Order LSP)
ATM
Switch
 IGP
SONET/SDH
ADM
FA-LSP
Ingress Node
(High Order LSP)
SONET/SDH
ADM
Egress Node
(Low Order LSP)
Egress Node
(High Order LSP)
attributes describing a forwarding adjacency
Local (ingress) and remote (egress) interface IP
addresses
Traffic engineering metric
Maximum reservable bandwidth
Unreserved bandwidth
Resource class/color (administrative groups)
Link multiplexing capability (packet, TDM, l , or fiber)
Path information (similar to an ERO)
46
ATM
Switch
LSP Hierarchy
PSC
Cloud
TDM
Cloud
LSC
Cloud
FSC Cloud
Fiber 1
Fiber n
LSC
Cloud
TDM
Cloud
l LSPs
Time-slot
LSPs
PSC
Cloud
Bundle
FA-PSC
FA-TDM
FA-LSC
Explicit
Label LSPs
Time-slot
LSPs
l LSPs
Fiber LSPs
(Multiplex Low-order LSPs)



Explicit
Label LSPs
(Demultiplex Low-order LSPs)
Nesting LSPs enhances system scalability
LSPs always start and terminate on similar interface types
LSP interface hierarchy
 Fiber Switch Capable (FSC)
Highest
 Lambda Switch Capable (LSC)
 TDM Capable
 Packet Switch Capable (PSC)
47
Lowest
Constraint-based Routing
Extended IGP
Routing Table
Traffic Engineering
Database (TED)
Constrained Shortest
Path First (CSPF)
 Reduces
the level of manual
configuration
 Input to CSPF
Explicit Route
 Path performance constraints
 Resource availability
 Topology information
RSVP Signaling
(including FA-LSPs)
 Output
 Explicit route for GMPLS signaling
48
User
Constraints
GMPLS Signaling Extensions
Label Related Formats (“Generalized Labels”)
 Generalized
label request
Link protection type (none, 1+1, 1:N, or ring)
LSP encoding type (packets, SONET, SDH, clear, DS-0, DS-1, …)
 Generalized
label object
Packet (explicit in-band labels)
Time slots (TDM)
Wavelengths (lambdas)
Space Division Multiplexing (fiber)
 Suggested
label
Label can be suggested by the upstream node
Speeds LSP setup times
 Label
set
Restrict range of labels selected by downstream nodes
Required in operational networks
49
GMPLS Signaling Extensions
PATH
SONET/SDH
ADM

RESV
SONET/SDH
ADM
Bi-directional LSPs
 Resource contention experienced by reciprocal LSP using
separate signaling sessions
 Simplifying failure restoration in the non-PSC case
 Lower setup latency

RSVP notification messages
 Notify message informs non-adjacent nodes of LSP events
 Notify-ACK message supports reliable delivery

Egress control
 Terminate LSP at a specific output interface of egress LSR
50
Link Management Protocol
LMP
 The
LMP
LMP
link between two nodes consists of
LMP
Control Channel
Bearer Channel
 An in-band or out-of-band control channel
 One or more bearer channels
 Link
Management Protocol (LMP)
Automates link provisioning and fault isolation
 Assumes the bi-directional control channel is always
available
 Control
channel is used to exchange
 Link provisioning and fault isolation messages (LMP)
 Path management and label distribution messages
(RSVP or CR-LDP)
 Topology information messages
(OSPF or IS-IS)
51
Link Management Protocol
Services Provided by LMP
 Control channel management
 Lightweight keep-alive mechanism (Hello protocol)
 Reacts to control channel failures
 Verify physical connectivity of bearer channels
 Ping test messages sent across each bearer channel
 Contains sender’s label [(fiber, λ) pair] object for channel
 Eliminates human cabling errors
 Link property correlation
 Maintains a list of local label to remote label mappings
 Maintains list of protection labels for each channel
 Fault isolation
 “Loss of light” is detected at the physical (optical) layer
 Operates across both opaque (DXC) and transparent (PXC) network
nodes
52
Link Bundling
Bundled Link 1
Bundled Link 2
 Multiple
parallel links between nodes can be advertised
as a single link into the IGP
Enhances IGP and traffic engineering scalability
 Component
links must have the same
Link type
Traffic engineering metric
Set of resource classes
Link multiplex capability (packet, TDM, λ, port)
bandwidth request)  (bandwidth of a component link)
 Link granularity can be as small as a λ
 (Max
53
GMPLS Benefits

Open standards allow selection of best-in-class equipment

Routers have visibility into the transmission
network topology
 Eliminates N2 meshes of links scaling issue
 Reduces routing protocol stress
 Optical paths span an intermix of routers and OXCs to deliver
provisioning-on-demand networking

Leverages operational experience with MPLS-TE

No need to reinvent a new class of control protocols

Promotes parallel evolution of UNI and NNI standards

Enables rapid development & deployment of new OXCs
54
GMPLS: Modern Thinking
for Modern Times


Aligns with the way that the next generation network needs
to be built and managed
20th Century – Transmission network was dominant
 Voice ran over the transmission network
 ATM/Frame Relay delivered private data services
 Internet was just one among many services
 Transmission network created subscriber services

21st Century – Internet is dominant
 Routers create the services that matter ($)
 Network must be optimized for IP/Internet
 OC-48/OC-192 make routers the largest consumers of bandwidth
 New architecture is driven by routers subsuming functions
previously performed by the transmission network

The transmission network must evolve in a way that is most
beneficial to the creation of Internet services
55
Thank You
http://www.juniper.net
Copyright © 2000, Juniper Networks, Inc.
56