sdn summit - optical sdn: virtualizing the transport network

Wes DoonanSDN Summit

March 2013

Optical SDN: Virtualizing the Transport Network

© 2013 ADVA Optical Networking. All rights reserved.22

Transport SDN

• SDN driving changes to client layer networks• Broadening access to the control plane

• Programmatic control interfaces to L2/L3 devices• New control paradigms for enterprises, datacenters• Separation of hardware/software functionality

• Leverage common software development models/practices• Utilize scale-out, datacenter-oriented architectures for control• Network-wide view of capabilities and states• Utilize view onto whole networks to drive efficiency and speed

• Client layer networks ride atop transport networks• Potential SDN benefits for transport

• Closer relationship between applications and networks• Hasten innovation, new approaches to network control• New applications, operational models, business opportunities

• Real SDN challenges• Strong provider/consumer network separation often required• Element complexity, technology complexity, OA&M complexity ...• Multiple data plane technologies, strong organizational barriers


Potential Use Cases

Bandwidth Calendaring Cloud-bursting

Secure Multi-tenancyWorkload Balancing

Transport networks increasingly asked to provide dynamic, high bandwidth, low latency services for SDN-enabled endpoints

Cloud DC

Private

Datacenters

Tenant 1

Tenant 2Load Load


Realities

• Transport networks largely serve IP networks today• Legacy voice networks still very much alive, but trend is clear

• Transport control can adapt to fit packet network needs• Transport networks already robustly model adaptations• Transport control planes familiar with managing complexity

• Transport networks exhibit specific data models• Key is to find the appropriate level of abstraction

• IP networks interested primarily in packet aspects• Packet traffic often engineered into explicit flows

• "Flow" entity in packet networks is thus a key abstraction

• What are the primary attributes of interest in a packet flow?• Set of connected endpoints• Committed bandwidth• Latency end-to-end• Fate sharing, recovery aspects

• These attributes point to the appropriate level of abstraction


Overlay Network Topologies

• How can transport networks support packet flows?• Simple connectivity no longer enough; richer model needed

• Endpoints are nodes in a topology; bandwidth and latency are attributes of links and nodes in a topology; fate sharing determined by the structure of a topology; recovery capabilities for a flow determined by the containing topology ...

Sensing a common thread here?

• Transport network virtualization via ONTs• Server network aspects expressed to client network in client terms

• Client network methods and techniques can remain unchanged• True for traditional EMS/NMS, distributed control plane, emerging SDN

• ONTs already much in use within client layer SDN today• ONTs achievable through common virtualization mechanisms


Virtualization

• Techniques for producing ONTs

• Node Scope Virtualization• Partition individual physical nodes into multiple virtual nodes• Represent individual complex nodes as groups of simple nodes

• Link Scope Virtualization• Represent paths in one network as links in other networks• Manage connectivity, fate-sharing, and adaptation

• Network Scope Virtualization• Represent networks as individual nodes in other networks• Manage scaling, information scoping, legacy integration

• Overlay Network Topology• Topologies containing mix of real and virtual components• Tailored to specific clients, applications, use cases• Express and subsume specific adaptations, policies, constraints


Node Scope

• Partition physical node into several virtual nodes• Assign physical resources to different virtual nodes• Split asymmetric physical node into multiple symmetric virtual nodes

Matrix:

1->4

2->5

3->6

A->7

B->8

C->9

OR

1

2

3

7

8

9

A B C

4 5 6

2 8

B

5

1

3

7

9

A C

4 6

1

2

3

7

8

9

A B C

4 5 6

1

2

3

4 5 6

7

8

9

A B C


Link Scope

• Paths in one network become links in another network• Squelch unneeded information (inline amplifiers, etc)• Subsume complex layer adaptations

• Example: path in ODUk network becomes link in Ethernet network

Server

Client


Adaptation Details• Transport network can provide client topology via adaptations

• Example: OTN transport element providing Ethernet link

10TCC10G-ADM 8ROADM-C80/0/OPM 10TCC10G-ADM8ROADM-C80/0/OPM

…

= Ethernet Link

= ODUk Trail

= OCh Trail

Packet

Transport

= Virtual Node


Network Scope

• Networks become virtual nodes in other networks• Respect administrative, security, regulatory boundaries• Encapsulate technology domains, legacy domains, opaque domains


Virtual Node Activation

C

D

C

D

C

D

"Create flow from C to D"

Application

NMS SDN (sub)Controller CP, PCE

"What happens in a VN, stays in a VN"


Transport Overlay #1

App#1 utilizes server network via specific network overlay

Application Application

= virtual node

= real node


Transport Overlay #2

App#2 utilizes same server network via different network overlay

ApplicationApplication

= virtual node

= real node


Overlay Activation

Applications activate overlays via SDN or existing NMS/CP

Application

Controller

= virtual node

= real node


Overlay Aspects of SDNOverlays are nothing new

• Node virtualization already available for packet devices• OF-Config/OVSDB enable segmentation of forwarding planes

• Convert one physical switch into multiple logical switches

• Path virtualization already embodied in packet networks• OF logical ports can terminate NVGRE/VXLAN encapsulation

• In effect, new packet links between non-physically adjacent nodes

• Network virtualization also embodied in packet SDN• Flowvisor concepts and components, slicing techniques• Other interesting uses for network virtualization

• Encapsulate legacy systems within a broader SDN framework

• So, are we done then? Nope• Transport introduces new aspects to SDN architectures


Asymmetric Switches

• Packet switching matrices generally symmetric• Any packet can be switched to any port

• Transport switching matrices often asymmetric• Multiple complex constraints on how traffic can be switched

• Example: OTN Hierarchy• Specific containment for ODU containers

• ODU0 -> ODU1 -> ODU2 -> ODU3 -> ODU4, ODU0 -> ODU2 -> ODU4, ODU2e -> ODU4, etc

• Example: Optical ROADMs• Optical elements have wide range of constraints

• Channel continuity (e.g. optically transparent nodes)• Fixed filter structures (e.g. endpoint fixed to specific network degree)• Regenerator diversity (e.g. some tunable, some fixed)• Endpoint diversity (e.g. may be fixed, tunable, switchable, combo)

• Asymmetries in transport inject constraints into client overlays


Example Asymmetries

Each transport node variant introduces different set of asymmetries into the overlay networks they support

(λN:DANY) (λN:DANY)(λN:DANY) (λN:DANY)

(λN:DN) ΤN (λN:DN) ΤN (λN:DN) ΤN(λN:DN) ΤANY

oo o

(λN:D1) (λN:D2)(λN:D1) (λN:D2)

(λN: ) o (λANY: ) (λN: ) o (λANY: ) (λN: ) o (λANY: )(λN: ) (λN: )

capabilities

constraints

= passthru = add/dropT = terminalλ = channel D = degree

Fixed OADM 2D ROADM 3D ROADM, Flexible A/D 4D ROADM, Fixed A/D

Å

Å

ÞÞÞÞ

Þ Þ Þ Þ


Lack of Payload Visibility

• Packet SDNs often utilize payload visibility for core operations• Topology discovery

• Capture/inject LLDP frames, process at controller

• Performance monitoring• Packet counters, per flow / per port

• Transport elements generally have no payload visibility• Payloads mapped into containers at edges, core sees only container• Transparent transport is a feature, not a bug

• 100G muxponder versus 100G routing blade

• Matching capabilities extremely limited• Match on port; nothing else

• Possible actions similarly limited• Switch one port to another; nothing else

• Different ways to monitor performance• Timing, signal quality, … ?

10TCC10G-ADM 8ROADM-C80/0/OPM


SDN Aspects

• Introduction of SDN unlikely to proceed linearly• Network operators rarely rip up and replace working systems• Transport networks have existing management systems

• Transport technologies are complex, rigid, rife with legacy• Client application access to physical elements often disallowed

• Overlay networks permeating L2/L3 architectures• Concepts well understood by practitioners, benefits clear

• Abstraction, mobility, scaling, usual reasons

• Virtualization is a useful enabler

• Initially introduce transport via client SDN framework• Model transport components as elements of the client network• Utilize virtualization techniques to implement the model

• Allow time for SDN to mature, benefits to emerge• Compelling applications in particular will be critical to success


Summary

• Transport networks can utilize/integrate with SDN techniques• Transport networks have complexities• A continuum of techniques needed to realize benefits

• Overlay network topologies initially bring transport into SDN• Providers use ONTs to enable SDNs for clients• "Topology-as-a-Service", "Just Enough Topology" models• Mediation to physical devices eases provider concerns• Seamless coordination amongst technologies and layers• Applications deal with flows, ONTs handle transport details

[email protected]

Thank you

IMPORTANT NOTICE

The content of this presentation is strictly confidential. ADVA Optical Networking is the exclusive owner or licensee of the content, material, and information in this presentation. Any reproduction, publication or reprint, in whole or in part, is strictly prohibited.

The information in this presentation may not be accurate, complete or up to date, and is provided without warranties or representations of any kind, either express or implied. ADVA Optical Networking shall not be responsible for and disclaims any liability for any loss or damages, including without limitation, direct, indirect, incidental, consequential and special damages, alleged to have been caused by or in connection with using and/or relying on the information contained in this presentation.

Copyright © for the entire content of this presentation: ADVA Optical Networking.

http://www.linkedin.com/groups?about=&gid=1194227&trk=anet_ug_grppro

http://twitter.com/ADVAOpticalNews

http://www.facebook.com/pages/ADVA-Optical-Networking/37630238931?ref=ts

http://blog.advaoptical.com/

sdn summit - optical sdn: virtualizing the transport network

Technology