Network Virtualisation for Packet Optical Networks

1
Network Virtualisation for Packet Optical Networks

• Adrian Farrel, Old Dog Consulting
• Steve West, Cyan Optics
• Daniel King, Old Dog Consulting

2
Agenda

• Commercial Objectives
• Overlay Networks
• Mesh Topologies
• Network Aggregation and Regrooming
• Packet Optical Networks
• The Packet Optical Transport Platform
• Network Planning
• Placement of network function
• Supporting Multi-client Networks
• Summary

3
Commercial Objectives

• Offer a wide range of client services and
  connectivity
• Carry legacy and packet services
• Support wholesale services (more than 50 of
  traffic for some operators)
• Allow rapid provisioning of new customer services
• Support changing traffic demands
• Allow addition of new customer sites
• Enable rapid protection at lower layers (closer
  to the fault)
• Leverage maximum revenue from transport networks
• Dont waste optical bandwidth
• Dont provision expensive equipment that isnt
  needed (just-in-time deployment)
• Get scaling benefits from switching traffic at
  lower layers
• Integrate and simplify operations and management
  of multiple technologies
• Use a single management point to operate multiple
  layers
• Provide consolidated multi-layer network
  visualization
• Dynamically (and independently) reconfigure
  network layers
• Change the way client connectivity is provided
• Reoptimise the transport network
• Invisible to client network
• Change the connectivity in the client network

4
Overlay Networks

• Networks constructed from equipment switching a
  single technology
• IP/MPLS, Ethernet, TDM, OTN, DWDM
• Network resources may be partitioned to support
  different customers or applications
• Internet Backhaul, Video Distribution, Wireless
  Operator, Customer VPN, and service classes
• Client-Server network relationships
• Client networks links provided by connections in
  a server layer
• Client and server networks are independent
• Client has no visibility or control of server
  layer resources
• Server layer is payload agnostic
• A server may support multiple clients (possibly
  using common resources)
• A client may use connectivity from multiple
  servers
• In this example we use an optical network (server
  layer) to create interconnections between
• routers for packet network services (client layer)

5
Network Topology

• Layered Networks
• Each layer has its own topology
• The topology of the physical network is defined
  by the available physical resources
• Higher layers have virtual links that tunnel
  through server layers.
• Mesh Networks
• Networks are made up of different topological
  constructs including spurs, rings, hub and spoke,
  partial mesh, and full mesh
• Spurs and rings are deployed in sparse transport
  networks, mesh in the core
• All topological constructs can be represented
  using a mesh
• The mesh is most sparse at the edge, and more
  fully meshed in the core
• Topology is often more fully meshed in the higher
  layers, and more sparse at the lower layers

6
Full Mesh Topologies

• Packet routers are connected by a full mesh of
  packet-layer links
• Packet-layer links are realised by dedicated
  paths through the optical core

• Disadvantages of full mesh network
• Waste of transport resources
• Under-use of dedicated resources
• n2 scaling issues
• Complexity of server layer planning and
  management
• Edge nodes need more server layer resources (line
  cards, lasers, etc.)
• Abstraction Trap Client has no idea of physical
  path
• Cost of client services is high
• Protection may not be real

• Advantages of full mesh network
• Direct, any-to-any connectivity
• Minimize delay in provisioning new client
  services
• Server layer treated as a set of logical links
• No worries about client connectivity
• Simplified client network management
• Redundant connections in case of failure

7
Partial Mesh Topologies

• Packet routers are connected by a partial mesh of
  packet-layer links
• Packet-layer links are realised by dedicated
  paths through the optical core
• Packet delivery may require routing at
  intermediate routers

• Disadvantages of partial mesh network
• Reduced CAPEX offset by increased planning and
  operational costs
• Network planning more sensitive to demand matrix
• Network operation requires traffic engineering
• Avoid link congestion
• Guarantee resource sharing
• Data paths are more complex
• Paths may become long
• Protection harder to guarantee
• Routers may become congestion points

• Advantages of partial mesh network
• Reduced cost
• Reduced number of transport resources
• Lower nodal degree at edges
• Fits better with sparse lower layer topologies

8
Network Aggregation

• Traffic from multiple sources is collected
  together
• Aggregated traffic is for the same set of
  destinations
• Edges are attached as spurs
• Connectivity in core network is a simpler mesh

• Advantages of aggregation
• High degree of edge-to-edge connectivity
• More efficient use of core resources
• Share server resources among multiple client
  overlay networks
• Bulk data forwarding/switching at lower layer
• Model may be reproduced at multiple technology
  layers
• Edge equipment cheaper and simpler

• Disadvantages of aggregation
• Aggregation points must perform routing
• MPLS tunneling can reduce complexity
• Additional equipment cost and complexity at
  aggregation points
• Complex to plan and optimize
• Traffic demand changes can break the model
• Protection and resiliency may be harder

9
Regrooming

• Some nodes within the core network capable of
  performing routing
• Allows traffic to be moved from one trunk to
  another
• Traffic can also be switched at the server layer
  (default behavior)

• Advantages of regrooming
• Simplify the core mesh
• Make even better use of core resources
• Retain high degree of edge-to-edge connectivity
• Continue to perform bulk data forwarding/switching
  at lower layer
• Model may be reproduced at multiple technology
  layers

• Disadvantages of regrooming
• Need more sophisticated (expensive) nodes within
  the core
• Positioning of regrooming nodes is a headache
• Network planning and operation significantly
  complex
• Need dynamic software assistance?

10
Packet Optical Networks

• Objectives
• Satisfy the commercial objectives
• Carry packet traffic over an optical core
• Integrate packet and optical networks
• Achieve all of the benefits of the mesh,
  aggregation, and regrooming techniques
• Minimize the disadvantages!
• Harmonized NMS, OSS, etc.
• Deploy packet optical nodes
• Capable of packet and optical function
• Switching of optical circuits
• Termination of optical circuits for local
  delivery and for packet processing
• Routing, aggregation, and grooming of packets
• Deployed in an optical mesh
• Used to build a virtual packet network
• Use multi-layer nodes
• Capable of switching, aggregation, and regrooming
  at multiple layers
• E.g. packet, TDM, and WDM
• Utilize sophisticated planning and management
  tools
• Placement of equipment within the network

11
The Packet Optical Transport Platform
Client Interfaces

• Fundamental component of future networks
• Strategically placed within the network to
  perform aggregation and regrooming
• Each node is capable of playing a role in
  multiple network layers
• Potential for pluggable line cards,
  switch-fabrics, adaptation modules, and grooming
• Means that planning is less rigid
• Truck-roll flexibility does not require fork-lift
  upgrades
• Increased flexibility makes planning potentially
  very complex
• Key issues
• What equipment to put at each site?
• How to plan the virtual network at each layer?
• How to aggregate and groom traffic?

Packet Router
XC
XC
12
Network Virtualization and Visualization

• Graphical Display Tools
• Enhance operator understanding by clearly
  displaying subsets of links
• Concurrently display data for multiple network
  layers
• Show dependencies and resource allocations across
  layers
• What-you-see-is-what-you-get environment improves
  confidence and reduces operator errors
• In-service experimentation to help plan changes
• What-if scenario trials
• Sophisticated tools already available

13
Network Planning

• Network planning is a multi-layer optimization
  process
• Demands need to be determined at each layer
• Network links at one layer are demands at the
  next layer
• Planning dictates
• Logical topology at each layer
• Physical topology at the lowest layers
• Placement of grooming and aggregation function
• Networks are designed using off-line planning
  tools
• Select equipment, fibre, and bandwidth based on
• Projected traffic growth.
• Service levels (availability, service delivery
  times)
• Total network cost (capital and operating costs)
• Online network planning
• Determine when to add, modify, and remove logical
  links
• Plan and reserve capacity for shared and
  path-disjoint protection
• Trigger just-in-time deployment of network
  hardware
• Deployment of cards, cross-connects, and
  adaptation functions

14
Placement of Grooming Function
B
D
A
C

• Old network requires 14 lambda hops
• Full mesh of connectivity between routers
• Insertion of regrooming function (at D) reduces
  this to 7 lambda hops
• If traffic load A-to-B grows, a separate lambda
  can be used and can be switched at D

D
B
A
C
15
Network Planning Virtual Network Topology

• Dynamic traffic engineering and path computation
  components
• Distributed TE Routing in each layer
• Path Computation Element (PCE)
• Virtual Network Topology Management (VNTM)
• Virtualization of network resources
• Virtual Network Topology (VNT) is a tool for
  service and transport aggregation
• Better utilize available resources to support
  more client layers and services
• VNT supports inter-layer network engineering
• The virtual network topology can be tuned based
  on client demands
• Multiple server networks may provide transport
  trunks to multiple clients
• Network reoptimization
• Significant savings may be possible if resource
  allocations are occasionally reoptimized
• Changes must be subject to policy or operator
  supervision
• Virtual link flapping is to be avoided as it
  would flap lower layer resources
• Reoptimization includes both intra-layer (traffic
  engineering) and inter-layer (virtual topology)
• Intra-layer reoptimization will be required more
  often
• The rate of reoptimization should be
  significantly lower at the lower network layers.

16
Multiple clients

• A scalable and flexible packet optical
  infrastructure supports multiple client networks
• Lambda service layer
• Layer 1 VPN
• Native Ethernet
• TDM
• IP / MPLS

17
Summary

• Commercial and operational trade-off between
  connectivity and aggregation
• Full mesh network
• Provides maximum flexibility for service delivery
• Does not scale and is wasteful of expensive
  resources
• Traffic aggregation
• Can be dynamic using changing traffic demands and
  new service deployment
• Harder to plan and needs more sophisticated
  equipment
• Cost-effective network operation
• Photonic network is the foundation for scalable
  bandwidth and switching flexibility
• Multi-layer switching, grooming, and aggregation
  at strategic nodes
• VNT provides a powerful tool for managing a
  multi-layer network
• Sophisticated planning software will be required
• Substantial new revenue streams from multiple
  client networks
• Not all client networks are packet networks
• Maybe packet optical is the wrong name!
• Introducing Integrated Optical Networks

18
Questions ?

• Feel free to send us questions
• adrian_at_olddog.co.uk
• steve.west_at_cyanoptics.com
• daniel_at_olddog.co.uk