Dynamic Lightpaths in R&E Networks July 17, 2007 Jeff Verrant

Agenda What is a Dynamic Light Path? And Why? Technology Requirements

The following slides describe, at a very high level, how Dynamic Services, works. Coordination across networks is critical. The advanced R&E networking infrastructure in the US can be thought of having three geographic scales, all of which need to work together to provide the community the capabilities it requires.

In addition to the national infrastructures, there are the regional or state-based networks--represented here in the abstract by yellow ovals.

..and the campus infrastructures--represented here by gray ovals--including those that provide connectivity to unique scientific instruments, such as radio telescopes.

Today, there the network infrastructure that serves these is Internet Protocol based, represented here by yellow lines, and blue router nodes.

But the community is also developing services that provide dynamic circuit (or “DCS”) networking capabilities. Those capabilities are being built into the new Internet2 network. In effect, this is a dedicated infrastructure parallel to the “traditional” Internet2 high-performance IP service infrastructure, built on the same fiber footprint.

Similarly, regional and campus networks--which have already made great progress in acquiring dedicated optical infrastructures--are expected to deploy infrastructures that extends the dynamic circuit services to the endopints that require them--here illustrated as a radio telescope and a computing resources needed to process the data produced by the radio telescope. The IP infrastructure is used to signal the dynamic circuit service infrastructure to set up dedicated channel. The channel can be made on demand or in advance, the circuit can last for minutes or weeks, and can be from 100 megabits per second and 10 gigabits per second.

A New Networking Model About a year and a half ago, some of the leading thinkers in advanced networking built a model for thinking about the networking needs these advanced applications have. One way to look at this is as this century’s version of the classic 7-layer OSI networking model. We call it the 3-D networking model. It describes the different kinds of networking services in terms of reliability along one axis, duration along another, and the classic OSI model layers 1-3 along the third.

A New Networking Model

A New Networking Model So a particular service might be long-term, reliable, and IP-based. In fact, this is what we think of today as the high-performance Internet backbones that serve most of the leading-edge R&E networking community. But with the optical-fiber based infrastructures the community is deploying, we also have the opportunity to provide short-term, wavelength-based experimental services, which might be just the thing for an impromptu network research project. And, of course, ideally we’d be able to provide anything in between.

Layers : Dynamic Connections are not just wavelengths Layer 1 : Wavelength Switching Map an application directly onto a 10G wavelength Dynamically provision and switch the lightpath Layer 2 : vlan / ethernet – gfp – vcg / sonet ( OTN ) 50M – 10G, increments of 50M Apply push / pop vlan switching No stranded bandwidth Dynamic provisioning, dynamic sizing Layer 2.5 - 3 : MPLS / ip switching

Wavelength Switching Any channel from any node, to any node Remote provisioning of wavelength route Simplifies maintenance activities Improve stability, reliability and traceability by removing “fingers” from the network Accelerates rollout velocity Quickly re-route existing service Simplifies network planning No stranded wavelengths Safegaurds upgrade capacity Extends life of network Limited need for accurate node-by-node capacity projections OpEx and CapEx savings

Layers : Dynamic Connections are not just wavelengths Layer 1 : Wavelength Switching Map an application directly onto a 10G wavelength Dynamically provision and switch the lightpath Layer 2 : vlan / ethernet – gfp – vcg / sonet ( OTN ) 50M – 10G, increments of 50M Apply push / pop vlan switching No stranded bandwidth Dynamic provisioning, dynamic sizing Layer 2.5 - 3 : MPLS / ip switching

Research Networks The Need for Flexible Lightpaths Connectivity Requirements Guaranteed Deterministic Bandwidth (10s Mbps – 10Gbps+) Scheduled-Demand Bandwidth; Hours, Days, Weeks Low Latency Data Replication Multi-site correlation High Availability Mulitple Communities of Interest 200Mb 10Mb 500Mb 1Gb (FC) 10Gb SDH / IP/ Ethernet Scarce Resources Data Collection Data Crunching Data Storage

Globalisation is a Reality Global, Multi-Domain Connectivity Service definition across Protocol boundaries Service Creation across domains Network Element Interworking functions 200Mb 10Mb 10Gb SDH SONET MPLS Data Collection Data Crunching Data Storage

Internet2 Dynamic Circuit Services (DCS) Image Slide I2 HOPI: Force10 E600 I2 DCS: Ciena CoreDirector 10 Gigabit Ethernet 10 Gigabit Ethernet 10 Gigabit Ethernet 1 Gigabit Ethernet OC192 SONET/SDH 1 Gigabit Ethernet or SONET/SDH

Integration : Core Director Domain into the End-to-End Signaling VLSR uni-subnet LSR upstream LSR downstream signaling flow uni, tl1 uni, tl1 data flow Ciena Region CD_a subnet signaling flow CD_z Signaling is performed in contiguous mode. Single RSVP signaling session (main session) for end-to-end circuit. Subnet path is created via a separate RSVP-UNI session (subnet session), similar to using SNMP/CLI to create VLAN on an Ethernet switch. The simplest case: one VLSR covers the whole UNI subnet. VLSR is both the source and destination UNI clients. This VLSR is control-plane ‘home VLSR’ for both CD_a and CD_z. UNI client is implemented as embedded module using KOM-RSVP API.

Graphical User Interface Monitoring and Control Ciena Core Director “NodeManager” Timeslot Map Network Utilization Monitor

Optical Transport Network (OTN) ITU Standards G.709 “Digital Wrapper”, G.872, G.873.1 Defines line/muxing rates, Optical Transport Unit (OTU) ODU-1/2/3 payload in OTU-1/2/3 = 2.5/2.7Gbps, 10/10.7Gbps, 40/43.0Gbps OTU-2 supports 10GbE LAN PHY (Extensions to include Preamble, Over-clocked for IFG) OTN & SONET/SDH share same foundation Similar framing with addition of OTN FEC Powerful OA&M capabilities (GCC0 akin to DCC) Asynchronous and Transparent Services with different clock sources integrated side-by-side Secure; Client OAM channels maintained Traffic Payload FAS: Frame Alignment Signal OTU: Optical Transport Unit ODU: Optical Data Unit OPU: Optical Payload Unit Overhead for OA&M Forward Error Correction

Evolution of the Client-Server Network IP Alien Wavelengths FC ESCON SONET/SDH is Managed Transport “Server” layer for existing service “clients” IP TDM Voice PL ATM SONET /SDH Ethernet IP IP builds over WDM … so does Ethernet WDM OTN OTN provides the necessary Managed Transparent Service for all Transport Clients … and ESCON, FC, l services WDM is now an unmanaged network “Server” to many transport “Clients” (which now includes SONET/SDH) WDM augments SONET/SDH capacity Animated Slide

Emergence of Connection Oriented Ethernet WDM Alien Wavelengths FC ESCON IP COE* Ethernet TDM Voice PL ATM SONET /SDH OTN Driven by Demand for packet focused replacement of SDH Robust as SDH Less Complex than MPLS Less Costly than either Connection oriented for deterministic B/W Disable MAC learning, Broadcast Unknown, STP Explicit Paths and CAC for guaranteed QoS and Restoration High Availability Transparent L2 Aggregation Mux Efficiency *COE: Connection-oriented Ethernet

Thank You!