Differentiation in the Age of Optical Convergence TNC 2019, Tallinn Rob Smets, Optical Transport Architect, SURFnet Jonathan Homa, Portfolio Marketing, ECI

Presentation Agenda SURFnet Practice Applications Framework Rob Smets - SURFnet Applications Framework Jonathan Homa - ECI

Introduction Services, NRENs, and Infrastructures Key Technological Developments Follow the Traffic Applications in a ROADM and CBF Network Infrastructures are resources, Services are demands on these resources, and NRENs are the poor souls that need to balance these.

Services and Infrastructure: A Delicate Balance The need for different type of services changes the role of NRENs: Traditional Telecom Operator DCI Provider (Secure) Carrier Niche Provider Alien Waves Spectrum Specials The transport services needed for these roles must rely on the same underlying infrastructure and control paradigms. This becomes a delicate balance. Smokestack

One Infrastructure to Rule (Transport) Them (Services) All! Optical Fiber Still single mode, one core per fiber. No MCF, FMF available. On-net deployment (relatively) fast, off-net deployment slow Very expensive in terms of correcting an imbalance Optical Light System 10-year technology cycle, 3-month topology change, seconds service change Medium expensive in terms of correcting an imbalance, but still complex to do. Transponders 2-year technology cycle, 1-week infrastructure, seconds for service changes. Medium expensive to correct an imbalance, if time and proper planning permits

New Technological Development Well documented and standardized optical, line, and control interfaces. Extremely high on the wish list! Needed by the automation layer FlexGrid and Contentionless ROADMs Programmable transponders: OTUCx, FlexO, 400GbE, FlexE, required throughput and allowed spectrum. Programmable OSNR margin to NL-Shannon limit. Interoperable transponders: oFEC, striping, and filtering Performance Monitoring: programmable data bins using streaming telemetry. Static resources become dynamic and programmable resources

Follow the Traffic! For any of these roles, best practice is to follow the traffic. In the short term: inefficient use of bandwidth is perfectly ok, if it agrees with installed capacity. In the long term: inefficient use of bandwidth is extremely expensive. Traffic matrices can change very rapidly. Service popularity Geographic divergence SURFnet and a large number of other NRENs acknowledge these developments. There is an ever important need to assure the balance not only on the long term, but also on the short term. We see huge increases in traffic to MS Azure and also in the new academic year, universities have improved their network, new education techniques have been implemented on open and online education, which consume more bandwidth. We also see more universities implementing their backup ICT facilities in a datacenter and that they require connectivity and bandwidth to this new dislocation. 5G on campus solutions for in-building coverage is also an example of an application that attracts more traffic that instead of over the Mobile Operator’s network will be transported over SURFnet’s network.

Classic WSON based restoration 100G, 50GHz, QPSK 100G, 50GHz, QPSK 100G, 50GHz, QpSK

Elastic and centralized-controlled restoration 200G, 75GHz, QPSK 200G, 62.5GHz, 8QAM 200G, 37.5GHz, 16QAM

Capacity optimization of a NL-GN constrained link G.655 fiber “lacks” dispersion to reduce NL-GN. Lowest ONSR values achieved in the blue part of the spectrum Links with shortest distances No complex modulation Use more spectrum Highest OSNR values achieved in the red part of the spectrum Longest distances Most complex modulation Use less spectrum

Capacity optimization of a NL-GN constrained links London Amsterdam Brussels Paris Geneva 8QAM 62.5GHZ Ultra-Long Long medium Short Make it from blue to red Voice over on the modulation format and spectrum C-BAND 1525nm 1535nm 1545nm 1555nm 1565nm

A Full Elastic Network and Flexible Restoration has huge benefits: Fiber layer: Delays the need for more fiber to become part of the core network Complex-modulated coherent-detected waves on new spans will be introduced by customer demand and not core transport demand. Easier network planning. Higher loading of Cross Border Fibers Partial restoration: Even if e.g. a 400G, 62.5GHz, 16QAM wave cannot be restored, restoring to 300G, 75GHz, 8QAM would still be useful and may trigger new fiber infrastructure. Postponing regeneration and topology impacting restoration

Presentation Agenda SURFnet Practice Applications Framework Rob Smets - SURFnet Applications Framework Jonathan Homa - ECI

Programmable Open Optical Networking SDN Control Open interfaces (Netconf/Yang) Open interfaces (Netconf/Yang) Programmable line rate Programmable client rate Programmable spectral grid 600G 300G 100G 100GE/400GE/FlexE Programmable lambda routing Programmable OTN switching

Programmable Open Optical Networking MAJOR FEATURES UNTIL NOW GOING FORWARD BENEFIT Openness Alien wavelengths Open interfaces (e.g. OpenROADM MSA) Modularity, disaggregation Control Monitoring NMS (procedural, manual) Alarms, events SDN (abstract, automated) Continuous telemetry Programmable control ROADMs OTN switching Fixed grid, colorless, directionless Medium/high capacity Flexible grid, Contentionless Access-to-core scalability Programmable routing Client side Line side Multi-service (Ethernet,storage,etc) OTUk, ODUflex 400GE, FlexE OTUCn, FlexO Programmable transport mapping Line rates Spectrum Coarse (100/200G) Fixed Fine (50G steps to 600G) Flexible Programmable throughput

Programmable Throughput Old Paradigm New Paradigm 50G steps to 600G Flexible grid in 12.5GHz steps with continuous fine tuning 100G 200G 50GHz 100GHz 50G 600G 37.5GHz 100GHz Line rate Channel width Line rate Channel width When we look back at old or existing paradigm in a few years, we will see how simple it is. Only a few line rates and a few channel widths. Not much choice, so inherently not efficient, or getting close to maximizing throughput. New paradigm is essentially continuous choice. Small 50GHz steps on line rate, and use of flexible grid for channel width. At least 70 hard options, with additional fine tuning. 70+ hard options plus fine tuning delivers programmable “continuous modulation” Maximizes throughput on any given link Essentially only 4 options

Continuous modulation envelope Hits “Sweet Spot” For any application Continuous modulation envelope 1000 3000 2000 4000 5000 6000 Line rate (Gbps) 7000 50 Distance (km) 600 Metro and Regional DCI ULH and Submarine 300 400 500 100 200 Long Haul 37.5 GHz 87.5 GHz Continuous modulation envelope

Programmable Transport Immediate Application Maximize throughput Old Paradigm Programmable Transport Traffic 200G link 5 x 1T 10 x 100G link 10 x 100G link 25 links 15 links with capacity to spare Traffic 350G link 3 x 1T 7 x 150G link 200G link 5 x Powerful applications, immediately and in the future. Immediate application is to maximize the throughput of a fiber, reducing he number of channels required to support a certain demand. Let’s look at an example. We have three nodes with demand of 1T between each of them. In the past we could only use 100G or 200G with 50GHz. Based on link engineering, this produced a best case of 25 links. With programmable transport, by playing with the modulation schemes and channel width we can do much better. Net result is an ability to carry same traffic on only 15 links! In high demand situations this sells itself.

New Application Dynamic restoration Programmatically re-arrange links to restore services due to failures Fiber cut on 300G @ 75GHz link carrying 3 x 100GE clients Programmatically finds and assigns a “longer” replacement link, carrying 300G @ 87.5GHz 2 1 300G @ 75GHz In these examples client traffic can be regular GE or FlexE. FlexE can also provide some programmability on the client side that can be coordinated with optical transport. But does not need to be FlexE clients. In first example, transport is carrying 300G of GE/FlexE traffic over a 75GHz link. Link is cut, can find a longer replacement link, but requires more bandwidth. No problem, up this to 75GHz. This just tip of iceberg. Same approach combining programmable transmission with programmable routing, makes for very powerful dynamic restoration. If FlexE, then can also apply sub-rating. For example, can only find 200G backup path, inform FlexE, can continue service at lower rate rather than lose altogether. 3 x 100GE/FlexE 2 300G @ 87.5GHz

New Application Bandwidth on demand Programmatically makes bandwidth available in response to customer requests Customer requests temporary activation of additional 100GE interfaces Programmatically assigns bandwidth within network 1 2 Platforms like Amazon Web Services let business customers order computing resources on-demand, raising their expectations for similar telecom offerings. For example, a data center operator may normally requires a 100G link between two locations, but on occasion, to perform large backups, would want to increase this to 300G. To accommodate this, the customer would purchase three 100GE physical interfaces from each data center, and a 100G transport service. The customer pays a fixed fee for this basic package. Through a portal, the customer can ‘dial this up’ to 200G or 300G whenever more bandwidth is needed, and the programmable network allocates resources to accommodate the request, or makes a best effort to do so. The rates may vary depending on time of day, and the customer just pays for the additional bandwidth during the temporary usage period. Basic 100G link 3 x 100GE/FlexE Grows to 200G or 300G link Three physical interfaces, with flexible service agreement: One is permanently active Other two available on demand

Future Application Just in time margin Old approach Programmable network On day 1, 50% of lambda capacity (and sometimes more) is unused for margin As fibers degrade over many years this margin is used up On day 1, nearly 100% of lambda capacity is allocated for traffic The network continually redistributes traffic as fibers degrade Many industries operate on the principle of “just in time inventory”. A car manufacturer will receive the parts it needs for the day’s production, perhaps only the previous day. This eliminates idle capital. Optical routes, however, typically are overbuilt by about 3dB to 5dB margin, to compensate for the future degradation of the fiber. This is like holding additional inventory for over 10 years. With “just in time margin”, the SP provisions the route right up to what it can handle. When the route does eventually deteriorate, the line rate can be restored in variety of ways, such as by increasing the channel bandwidth, or provisioning a partially new route (like in diagram), ir brand new route and releasing the existing facilities for other services. In other words, the network continually optimizes itself. When the network controller (an ML/AI algo in the future) sees there is no more possibility to allocate new bandwidth, it “orders” more capacity to be installed. $ $ $ $

Disaggregation Approach Netconf/ Yang Open programmable OTN switching system OTN Switching Netconf/ Yang OTN Transport Transponders & Muxponders Open programmable transponder system Independent sub-system deployment from multiple vendors Based on interoperability specs from organizations like OpenROADM MSA Enables best-of-breed for different functions Promotes differentiation of optical networking solutions Open programmable optical line system Line system (CDCF ROADMs & Amps) Netconf/ Yang 600G 300G 100G 100GE/400GE/FlexE

Apollo Optical Networking Programmable Maximizes capacity and responsiveness Open Enables multi-vendor disaggregated solutions Summary of Apollo value proposition Until now, and continues to be: low cost per bit, easy to operate, with value added features like encryption, protection, and restoration. Now also a truly programmable and open platform, can be used to create a more dynamic and responsive network, bringing new service, and cost efficiencies. Also, an ability to assemble Apollo capabilities along with other vendor capabilities, “as you like it”. 23

Thank You for Your Attention!