Transport SDN: The What, How and the Future!
Inder Monga Chief Technologist & Area Lead, ESnet ONF Research Associate
Transport SDN Panel
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
Agenda Topics
What is transport SDN?
Use-cases
Transport SDN demonstration
Architecture and Standards
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
What is Transport SDN?
Control of Optical Transport networks using OpenFlow
Network elements expose the switching, cross-connect and flow aggregation capability
• L0 and L1 switching and cross-connects • Hybrid packet-optical switching and aggregation
Panel discussion today will expose multiple facets of this new direction
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
Transport SDN Block Architecture
Network Infrastructure Layer
SDN Controller
OFw and Ofc Topology, control & statistics
OFw: OpenFlow Wire OFc: OpenFlow Config
Network Control Layer Network
Virtualization
SDN Controller SDN Client
OFw and Ofc
Applications
Northbound Interface
Northbound Interface
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
Agenda Topics
What is transport SDN?
Use-cases
Transport SDN demonstration
Architecture and Standards
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
R&E network design trends driven by ‘Big-data’ Flows…
Buying Circuit Services
Lease Dark Fiber
Multiple 10G (Multiple) 100G
IP Dialtone Steer ‘Big Data’ flows
Regional Footprint
Global, Federated Multi-domain
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
…drives new network architectures (1) • [Dynamic] Optical Bypass
• Separation of large flows, from L2/MPLS circuits to L1 circuits
• Traditionally has been very complicated, vendor dependent and not-interoperable
• ARCHSTONE Extensions • Multi-Layer Topology Representations • Multi-Layer Topology Computations • Multi-Layer Provisioning
• Layer Decisions • driven by resource constraints or
client requests for specific performance characteristics (i.e. low latency, low jitter, etc)
ARCHSTONE project DOE
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
…drives new network architectures (2) • Packet-Optical Integration driving layer collapse
• Packet switching “marries” Optical transport nodes • thanks to merchant silicon
• LH Optics “marries” core routers • thanks to coherent technology and optical integration
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
…drives new network architectures (3)
• No uniform way to manage or create multi-vendor, multi-layer networks or multi-domain optical bypass
• Packet and Optical devices managed in totally different ways
• Multi-vendor networks are islands
• No global topological visibility • GMPLS and vendor-specific hacks needed in current approach
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
ESnet Transport SDN Demo
bnl-tb-wdm-3 bnl-tb-wdm-4
40G
100G
20G 20G
20G L1 Tunnel
SDN Controller communicating with OTS via OpenFlow extensions Bandwidth on Demand application for Big Data RDMA transport 3 physical transport path options (with varying latencies) Implicit & explicit provisioning of 10GbE/40GbE services demonstrated
Topology Monitoring App BW on Demand App
ESnet SDN Controller
Mellanox Mellanox
Path #1
Path #2
Path #3
OTS
ESnet LIMAN Production Network
Brookhaven National Laboratory Testbed
OTS
OSCARS
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
Community has taken it on to build Standards
Open Network Foundation has formalized a Open Transport Working Group
• Participation from many transport, optical and router vendors
OIF has formed a ‘Carrier SDN’ group • To discuss carrier requirements for SDN
Other conversations and standards body co-ordination in place
Lawrence Berkeley National Laboratory U.S. Department of Energy | Office of Science
Why Transport SDN?
Automation • Centralized management applications aka OSCARS
Scaling • End-to-end service delivery • Works for IP and non-IP protocols
Virtualization • Multi-layer abstracted to a single end-to-end connection
12/18/12 Inder Monga, OTS Demo 12
Top Related