Cisco PONC 2015 · • Cisco and Third Party Applications. Leveraging REST APIs, interface to...
Transcript of Cisco PONC 2015 · • Cisco and Third Party Applications. Leveraging REST APIs, interface to...
Cisco PONC 2015
Walid Wakim
Principal Engineer
March 10, 2015
2 © 2013-2014 Cisco and/or its affiliates. All rights reserved.
Transport SDN
Network Objectives
• Today
• Network Planning performed in silos
• Network Architectures are viewed in Layers
• Operations are Layer Focused
• When talking about Optimizing it is not about the network but the layer
• Goal
• Simplify and Streamline Operational Models – Global View and Global Impact
• Improved SLAs enhancing the User Experience – “Best Effort”??
• Reduced Capital Expenses based on total network optimization - >60% savings
• Increase Service Velocity – Months to Minutes
• Allow multi-vendor / multi-domain operation – From a single controller
Growth Trends (Cisco VNI)
4Billion Internet Users
52% of world Population
Faster Broadband Speeds
2.6 fold increase in user rates
More Connected Devices
21Billion devices
Video continues to dominate
79% of all internet traffic
Mobile connectivity
50% of all connections are mobile
Emerging Trends
Focus on Service Optimization
Rather than layers / elements
Step-Up Network Convergence
New Multi-Layer Opportunities
Dynamic Service Activation
Anywhere, anytime
Static to Dynamic Transport
Flexible data rates and spectrum
Dynamic = Complexity?
SW - Simplify, Simplify, Simplify
Market Trends – 2018 Projections
Router
Optical Optical Domain Routing Domain
Abstraction View SDN Controller
Open Flow Provisioning
Configuration
Technology
Capabilities
Etc.
Provisi
oning
App
Manag
ement
App
What is Transport SDN • Utilizing Centralized SDN controller
to control Optical Transport networks
• Network elements expose Switching, Xconn, BW. Aggregation etc… to controller
• OpenFlow appears to be the interface of choice
• Application drives controller to act on Optical Transport network
This is very useful… BUT, is it enough??
Cisco View:
• Layered SDN in not enough
• Treat entire network as ONE is key
• Focus on Service, NEs are
Resources
• Flexible NE interfacing is required
• Hybrid Control is Key
• A Trusted path to the End Goal is
important
• Netconf Yang would be preferred
DWDM
OTN
Path Forward • Focus on Service
• Network Layering must be Transparent to the Service
• The Network must behave as a single entity not multiple layers – A pool of resources
• The Network Shall Support Service Activation Anywhere and Anytime
Architecting an Agile, Service focused network…
6
Service
Request
Service
Fulfilled
Layer-Agnostic Resource Pool
Packet
A B Service
Request
Service
Fulfilled
A B
UnifiedMLController
NetworkCollec on/DeploymentPlug-Ins
NetworkApplica ons
How do we Simplify
• ML Hybrid SDN Architecture is based on:
• Centralized Control for Optimization and Global views
• Distributed Control for Fast Reaction to Network issues
• Application Space:
• Cisco and Third Party Applications. Leveraging REST APIs, interface to Orchestration layer or directly to Controller
• Orchestration Layer:
• Service Orchestration across domains or a number of controllers. Cisco solution will also allow for direct device deployment
• Controller Layer:
• Unified Multi Layer Platform. WAN Controller, vendor agnostic, acting on all layers of the -> L0 to L3
Path to Cisco’s ML Hybrid SDN Architecture
How do we Simplify
• Multiple views in the industry on SDN architectures
• Many talk openness but require proprietary controller from specific vendor! Open???
Path forward
8
VendorandlayerSpecificEMSs
Packet'
OTN'
DWDM'
IP/MPLS'
GMPLS'
WSON'
Service' Service'
Packet'NE'
OTN'NE'
DWDM'NE'
VendorandlayerSpecificOpera ons,PlanningandDesigntools
3rdpartyorProviderbuiltOSS
Vendor‘A’EMS Vendor‘B’EMS Vendor‘Z’EMS
OSS
VendorandlayerSpecificEMSs
Packet'
OTN'
DWDM'
IP/MPLS'
GMPLS'
WSON'
Service' Service'
Packet'NE'
OTN'NE'
DWDM'NE'
Mul LayerController/Orchestrator
Vendor/3rdPartyApplica ons
Vendor‘A’Controller Vendor‘Z’Controller
MLController
Openinterfaces
Cisco’UnifiedMul LayerController
Packet'
OTN'
DWDM'
IP/MPLS'
GMPLS'
WSON'
Service' Service'
Packet'NE'
OTN'NE'
DWDM'NE'
Cisco/3rdPartyApplica ons
UnifiedMLController
Present Mode Possible Path Cisco Unified Multi Layer
How do we Simplify
• As Providers embark on SDN journey two paths will be available:
1. PMO directly to a full SDN Architecture
2. A phased path to build trust
Transitional Path to SDN
9
Rou ngDomain
DWDMDomain
• Independent IP/MPLS CP
• Independent Optical CP – WSON
• Wall separating layers
• No real information sharing
PMO
• Online Data Collection
• Multi Layer Co-ordination
• Multi Layer Feasibility / Restoration
• Online or manual Config
• nLight Control Plane Architecture
• Vendor Agnostic
Network Optimization Server • Remove the Wall
• Centralize CP - Global View
/Optimization
• Leverage Layered CP – Fast
Reaction
• Application Driven
• Vendor Agnostic
SDN
CLI/TL1/SNMP/NetConfUNI..
OF/PCEP/I2RS/TL-1/UNI
UnifiedController
OpenAPIs
PlugIn
BWCalendaring
orNOS
PrimeCarrierManagement
OpenAPIs
PacketLayer
Op calLayer
x
Op onal:PushConfignLight
CentralCompute
NetworkCollec on/“Deployment”
NetworkOp miza on
Server
nLightERO
WSON
IP/MPLS
Option push config with nLight
How do we Simplify
• Sharing of Relevant information:
Server to Client Client to Server
Latency Matching Circuit
SRLG Disjoint Circuit
Circuit ID LSP Priority
Path Restoration requirements
Cost Latency Bound
• Common Interest points
• Maintaining network data within the network
• ML Restoration saving of up 60%
• Protect against Multiple Failures
• Reduced Operation cycles
• Feasibility performed per circuit
nLigth Advantages
10
Client: IP layer
Server: DWDM layer
San Jose
LA
Seattle
Denver
DallasOrlando
Atlanta
Chicago
Ashburn
New York
25 Spans
2421Km17 Spans
1485Km
22 Spans
2090Km
6 Spans
682Km
30 Spans
2608Km
22 Spans
2097Km
13 Spans
1235Km
25 Spans
2159Km
9 Spans
772Km
13 Spans
1227Km
22 Spans
1852Km
5 Spans
460Km
15 Spans
1310Km
19 Spans
1780Km
Red Lines = Assumed Fiber
Black Lines = Real Fiber Sample
San Jose
LA
Seattle
Denver
DallasOrlando
Atlanta
Chicago
Ashburn
New York
Elk
(corp)
Cup
(corp)
Maiden
Miami
St Paul
Reno
Newark
Prineville
Corp PE
P
DC PE
Peering PE
Boston
nLight CP
Client
Server
Client
Server
How do we Simplify Network Optimization Server
11
L3 / L1 Tool
DWDM Tool
NOS – Network Optimization Server
• Multi Layer Planning /
Optimization Tool
• Online or Offline
• Time frame of weeks / months /
years
PacketLayer
Op calLayer
x
UnifiedMLController
NetworkCollec on/DeploymentPlug-Ins
NetworkOp miza on
Server
nLightERO
WSON
IP/MPLS
How do we Simplify
• ML network collection online
• Topology
• Circuits
• Resources
• Offline Network Analysis
• Impact Analysis
• What if Scenarios
• ML Restoration feasibility
• ML Optimization
• Coordinated Maintenance Feasibility
• Online Network Config or user config
• Vendor Agnostic leveraging Industry Proven tools and algorithms
Network Optimization Server
12
Packet'
OTN'
DWDM'
IP/MPLS'
GMPLS'
WSON'
Service' Service'
Packet'NE'
OTN'NE'
DWDM'NE'
NetworkCollec on/DeploymentPlug-Ins
NetworkOp miza on
Server
The SDN Network Architecture
• The Cisco ML Hybrid SDN Architecture:
• Application Layer – Cisco or 3rd Party SW apps
• Orchestrator – Orchestrates between domain controllers
• Controller Layer – Unified ML Controller, vendor and layer agnostic
• Network Elements – IP, OTN and DWDM elements.
• Cisco Architecture shall leverage:
• Centralized Controller for Optimizations and Activation
• Distributed Controller for fast reaction to local events
• Cisco Solution is robust operates with Cisco Controller as well as 3rd party controllers
• ML Applications become key
ML Visualization and Activation
Assurance
OpenDaylight REST API
Collector Network
Programming
Basic
Service
Inventory Topology Policy
Management
Analytics / stats
Modeler Carrier Ethernet
Data Center Specific
Service
SNMP NetConf PCEP OF 1.3
OTN$Domain$
Op, cal$Domain$
Rou, ng$Domain$
Network Applications
Unified Controller
Southbound Plugins
Packet'
OTN'
DWDM'
IP/MPLS'
GMPLS'
WSON'
Service' Service'
Packet'NE'
OTN'NE'
DWDM'NE'
Network(Applica/ ons( 3rd$Party$Applica. on$ Cisco&Applica+on&
Orchestra) on,Cisco&or&Third&Party&
Orchestrator&
SDN Architecture - Applications
• Multiple Applications being developed and investigated
• Applications shall operate over Cisco and Third Party controller
• Focus not simply on single layer but Multi Layer
• Partnership in place with third party for app Dev
• Initial Application will provide for ML:
• Visualization
• Service Activation
• ML Optimization
• Multi Layer Restoration
Applica on Descrip on
Mul -LayerVisualiza on TopologyandResources
Mul LayerProvisioning SimplesinglestepMLserviceac va on
Design/Planning GreenfieldorBrownfield,whatif,Strategic,etc…
CustomerPortal CustomerofthecustomerBusinessAppedtobillingandnetwork
Mul LayerOp miza on Op maluseofresources,periodicanalysis
Maintenance AlarmCorrela on,TCAsPMs,preemp veanalysisetc..
Mul LayerAssurance SLAs,QOS,Reports,etc…
SWManagement SWreleasemanagement
API to Controller
Cisco is Investing in ML Applications and Support Third Party apps
SDN Architecture - Controller Platform
• Built on key proven technologies / Components
• Modeling– Multi Layer Modeling built on Cariden acquisition as well as Cisco proven CTP and ODE
• Collection and Deployer – Cisco’s Open SDN Controller built on ODL implementation
• Collection and Deployer– Collector and Deployer of Legacy / non ODL NEs built on Cisco’s Tail-F acquisition
• Communicate between components leveraging JMS / REST with YANG models
NB RESTful APIs to Applications
Unified Multi-Layer Controller
Modeling (What-if)
Collection / Deployer / Services / Abstraction
Collector /
Deployer (non ODL
NE)
Southbound Plug-in to NEs
API API
APIs
Caride
n
(WAE)
Tail-F OSC (ODL)
Application Space
(Cisco / 3rd Party)
RFS Orchestrator
(Cisco Tail F NSC)
SDN Architecture - Controller Platform
• Built OpenDaylight Framework
• Opendaylight.org
• Cisco is a founding member
• Open Platform for Network Programmability
• Open sourced community
• 40 community members
• Leverage KARAF containers
• Lightweight OSGI runtime
• Provides container where different apps can run
• Ability to plug and play different apps
OpenDaylight (ODL)
Cisco Contributions Cisco contribution
SDN Architecture - Controller Platform
• Cisco ODL
• Taking Open Source and making it Carrier Grade
• Cisco is Largest contributor to ODL
• Plug-in approach allow for flexible interop
• MD-SAL layer – Model Driven Service Abstraction Layer
• Provides infrastructure / plumbing between consumer and providers
• Manages Plug-ins
• Provides common Abstraction
• Auto-generate NB APIs
Open SDN Controller – Cisco OSC
MD-SAL
OF NetConf PCEP BGP OVSDB
Base Network Services Domain Specific Services
Applications
Network
Topology
Inventory FRM
GBP L2 Switch
Controller
SDN Architecture - Controller Platform
• Southbound interfacing to Network elements
• Flexible and Modular solution
• Leverage ODL SB plug-ins
• Leverage Integrated Cisco Tail-F
• Directly config Network elements
• Provides immediate support for Legacy Equipment
• As interfaces move to ODL – simple plugin added
Interfacing to Legacy / non ODL NEs
Cisco OSC
SB Plug-ins
Network
Cisco
Tail-f
API
Leverage NEDs
Topology
MD-SAL
RFS Orchestrator
SDN Architecture - Controller Platform
• Service Manager
• Service definition
• Creates Specific configurations for service
• Device Manager
• Manages Device configuration
• Transaction based
• Real Time device changes
• Supports / Aware of external config source as well
• Configuration Data-Store
• Stores all config and Model data
• Maintains synch
Cisco Tail-F NCS Platform
Service Models
DeviceModels
Network Element Drivers (NEDS)
Device Manager
Service Manager
REST, NETCONF, JAVA
WEB UI, Network wide CLI
• NEDs – Network Element Drivers
• Provides the Device model
Defines configuration and attributes in YANG
• Mapping the devices native Interface
• Extensive library of multi vendor devices defined
SDN Architecture - Controller Platform
• Real Time Analysis
• What if
• Impact Analysis
• Strategic Analysis
• Feasibility
• Acts on all or single network layers
• Provides true Network Optimization
• Vendor Agnostic
• Even at the Optical Level!!
Modeling
Controller
Cisco OSC
SB Plug-ins
Network
Cisco Tail-f
NCS
API
Leverage NEDs
Topology
MD-SAL CDS
NEDS
D
M
SM
Modeling
API
SDN Architecture - Controller Platform
• Cross Layer Path Computation Execution (WAE):
• Built on Industry leading Mate Design Engine from Cariden Industry proven algorithms
• The brain for optimizing each layer is modularized as plug n play OSGI bundles and feeds feasibility and weight correlation of each layer to the central algorithm
• The central cross-layer optimization algorithm is part of the WAE engine which optimizes all layers to provide the least number of interfaces respecting all layer SLAs
Modeling – Cisco Solution
21
KARAF Container
OSC TAIL-F
Packet'
OTN'
DWDM'
IP/MPLS'
GMPLS'
WSON'
Service' Service'
Packet'NE'
OTN'NE'
DWDM'NE'
Plan Current Model
New Model
Cross-Layer Optimization and Prediction Algorithm
L3 Optimization and Prediction Algorithm
MATE Engine
OTN Optimization and Prediction Algorithm
ODE Engine
DWDM Optimization and Prediction Algorithm
CTP Engine
OSGI
WAE Engine
Recap
• Dynamic Networks will be a reality!
• Open ML Solutions will be key instruments in the success of future netowrks
• SW will provide needed simplifications
• Providers need to trust the network will behave as expected
• Two paths will be available
• Direct path to SDN
• Intermediate step with nLight
• What are some of the benefits of the defined Architectures?
2
2
Use Cases Multi-Layer Network Optimization Global network view | Optimization across layers
15% interface savings
Multi-Layer Service Activation Months to Minutes | Simple, focused applications
Constraint-based routing
Multi-Layer Restoration >40% Interface Savings | Zero Touches
Re-use stranded network assets
Coordinated Maintenance Multi-layer service awareness | Months to Minutes
Hitless multi-layer re-route
Use Cases Multi-Layer Network Optimization Global network view | Optimization across layers
15% interface savings
Multi-Layer Service Activation Months to Minutes | Simple, focused applications
Constraint-based routing
Multi-Layer Restoration >40% Interface Savings | Zero Touches
Re-use stranded network assets
Coordinated Maintenance Multi-layer service awareness | Months to Minutes
Hitless multi-layer re-route 0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
DTyear1 DTyear5 TEFyear1 TEFyear5
Baseline
MLBO
MLBO+MLR-O
MLBO+MLR-O+MLR-P
IEEE Communication Magazine Jan-Feb 2014
~60% interface
savings
Summary Not about individual Components nor features but the Service!
Coherent
CCOFS-ROADM
Form Factor & Power
Optical
Scale
Packet Processing
Form Factor / Power
IP
Prime
Multi Domain
Multi Platform
Management
IP / MPLS planning
Optical Design/Plan
Network Modeling
Planning / Design
Bring Layers together
Unified Management
Single Point of Control
Service Focused
Orchestration
ONE Network