cable metro packet optical transport
DESCRIPTION
At the Society of Cable Telecommunications Engineers Expo 2014, Andy Smith of Juniper Networks presented Juniper’s vision and architecture for a cable oriented packet optical core and metro transport system. Access insights and network diagrams in his presentation and learn more in his blog post: http://juni.pr/1rwapCG.TRANSCRIPT
Cable Metro
Packet Optical Transport
Juniper Networks
Distinguished Engineer & Chief Architect for
Cable MSO Networks
Andrew Smith
The Network is FundamentalBut there are too many manual, fragmented
parts
Manually Operated
Fragmented Domains
The Network Must:
• Remove barriers
• Synchronize dynamically
• Respond to accelerating change
IP TRANSPORTSERVICES
Networks Need to CustomizeCable networks need to adapt to changing
requirements and conditions
COMPLEX
Exponential Growth,
Fragmented,
and Manual
OPERATIONS
INEFFICIENT
Over-provisioned
and Hardware
Dominated
INFRASTRUCTURE
RIGID
Limited Analytics,
Fixed Policies,
Months to Change
SERVICES
Packet Transport is KeyDynamic, efficient, flexible, programmable
EFFICIENT
Any to Any
Low Latency
Dynamic
PACKET
SWITCHING
ELASTIC
Load Adaptable
Resource Optimized
Highly Available
MPLS TRAFFIC
ENGINEERING
PROGRAMMABLE
Intelligence,
Programmability &
Abstraction
APPLICATION
AWARE
Silicon – Heart of Modern RoutersCapacity exceeding Moore’s Law
Total single system capacity will exceed 100Tbps shortly
Switching latency, port-to-port, < 10 usec
640G
1.6T
3.2T
24T
2002
Single System n-Degree Capacity
100+T
2017
Moore’s Law
Power/Thermal Efficiency
Power Utilization(Tb per KW)
1.2
Thermal Emission (BTU per Gbps)
2.8
Capacity Footprint
Per Rack Capacity (Tbps) 7.68
Capacity Footprint (Gbps per cu-in)
533
Platform Characteristics
Depth (mm) 270
Minimum Power (KW) 1.2
Latency (us) 5
PTX3000 Reference Platform
Packet System EfficiencyElasticity and Efficiency are Mandatory
Statistical multiplexing – resource sharing and adaptation to service
demands
MPLS LSP Bandwidth reservations dynamically adapt to the constantly
changing network demands
Persistent and ephemeral path creation via SDN controller
Dynamic and planned adaptation via SDN controller
Packet Forwarding Engine
Efficiency
Elasticity
Reconfigurability
Peak value P1
Bit
-rat
e
time
Peak value P2B
it-r
ate
time
PE1 to PE2 traffic
PE3 to PE4 traffic
OTN circuit size
OTN circuit size
Wasted Bandwidth
Wasted Bandwidth
PE1 PE2
PE3 PE4
Core
B/W ?
Packet ElasticityStatistical Multiplexing
With packets, required bandwidth in
core is less than P1 + P2– Statmux gain is more pronounced with greater
flows and higher variation of each flow. Modern
cable networks exhibit this behavior.
– In contrast, circuit-switched networks require
one circuit of size P1 (at least) and another of
size P2 (at least)
TDM circuit sizes are quantized with
large steps, bandwidth is wasted by
‘rounding up’– OTN granularity is relatively very course. ODU0
– 1.25 Gb/s,
Cable networks are also very
asymmetric!
A Packet Transport MPLS LSPElastic and Adaptive, Service Oriented
B/W reservation & guarantees
ElasticAdaptive
Loss & delay reporting
Unidirectional
Bi-directional & associated
Facility protection
Path protection
Node protection
Control-plane separation
Multi-path aware
Point-to-point
Point-to-multi-point
QoS
OAM reporting
Hierarchical LSPs
Multi-protocol
Explicit or dynamic or loose hops
Multi-layer Meshed Network
Collapsed Packet-Transport Network Fabric
Modern Cable TransportPacket Applications, Services, Transport
Cable networks are dominated by
router-to-router IP services and data
center interconnect– Point-to-point DWDM designs
– Higher speed router links (Nx100G)
A response function of the network– Analytics at every transport node
– Offered load is visible at the transport layer
– Network fabric can react
Transport modern cable services– Full suite of contemporary high-speed
residential Internet services, IPv4 / IPv6
– Emerging cloud and multimedia services
– Commercial L2/L3VPN, Internet
– Cell tower backhaul
– Enterprise services
Peer Discovery
Property Exchange
1Tb Transport-group
800G Transport-group
Automate for SimplicityOptical Plug-n-Play: Dynamic DWDM
Optical Peers– Dynamic discovery of adjacent peers with
optical capabilities – 100G coherent
transponder
– Via supervisory channel or DCN
Optical properties exchanged– Session establishment if and only if optical
properties of physical link are interoperable
– Exchange of “transport group” characteristics
Transport-groups– Definition and exchange of desired bandwidth,
inclusive of wavelength restrictions
– Wavelength channel allocation
MPLS Transport Service CreationMerge Benefits of Packet & Transport
• LSP Control, Creation, & Path Optimization
• Path Diversity (Link, Node, Facility)
• Bandwidth Scheduling & Calendaring
• Fast Reroute Planning
• Programmable Path Cost Functions
• Optimized Exit Control
• Global Concurrent Optimization
• Container LSP association (Auto-B/W & LSP multi-path/load-balancing)
ROUTING Distributed control Stat mux gain Multi-hop resilience Service integration Feature breadth
TRANSPORT Centralized control Predictability Dedicated bandwidth Fast recovery Operational
simplicity
SDN TE
NETWORK
CONTROLLER
SOFTWARE-DRIVEN POLICY
Topology Discovery Path Computation Path Installation
PCEP - LSP discovery
IGP-TE, BGP-LS - TED discovery
jVision – Streaming Analytics
PCEP – Control/Create traffic engineered LSP
Netconf/YANGMay include: BGP, DMI, OpenFlow, I2RS
ANALYZE OPTIMIZE VIRTUALIZE
Routing Netconf/YANGPCEPJunos CSPF Algorithms
Self Optimizing EfficiencyStateful WAN Controller Creates Paths
Bin-Packing
Defrag
Premium Paths
Scheduling
Calendaring
Predictability
Adaptive TE
Inter-Domain Routing
Global concurrent optimization
Network lifecycle management
AgentRPD
Kernel
PFE PFE PFE
Sensors
Programmable objects to tap into any state of the network node
Capabilities can be enhanced over time
Standard IP based interface with data store
Decoupling
Telemetry data is outside JUNOS
Extendible client interface
Different analytic views can be provided independent of JUNOS
PCE Controller
S S S
Analytics
Leverage Deep AnalyticsMultiple sources of data for optimization
Jvision – provides analytical insight into Juniper products. Data ‘pushed’
to external nodes in band
Other sources of intelligence via PCE plugin -- real time video
application feedback for cable networks
In ConclusionOpen standards, interop and innovation
AUTOMATE
Open multi-vendor
technologies for
simplicity and agility
SCALE
Ultra-high
performance and
efficient systems and
solutions
OPEN STANDARDS
Open, multi-layer,
multi-domain SDN
and comprehensive
NFV