The Big Takeaways from OFC 2022 - www.infinera.com
contact button

The Big Takeaways from OFC 2022

March 18, 2022
By Tim Doiron
Sr. Director, Solution Marketing

A triumphant return to in-person events for the optical industry

After two long years, the optical industry gathered in person at OFC and the Optica Executive Forum in San Diego March 5-9. We had great meetings with customers, partners, suppliers, analysts, and investors, all eager to see demos and talk to us about our latest technology. Overall, I felt like this year’s show was more about incremental improvements and technical advancements and less about earth-shattering new product introductions.

With that as the backdrop, here are my top high-level takeaways from the event, followed by eight more detailed takeaways about solutions and technology at the event.

  • Supply chain remains top of mind for many service providers while causing frustration and creating lead time challenges. Most analysts expect that we will see some relief in 2022 and a significant improvement in 12 to 18 months.
  • The state of the industry is strong, with capacity demand nearly doubling every two years, increasing deployments of 5G wireless networks, and a move toward augmented and virtual reality.
  • Coherent optical engines will include both pluggable and embedded form factors through at least two more generational development cycles. While pluggable engines are making great strides in performance and capacity-reach, embedded engines continue to deliver the highest spectral efficiency and performance at long-haul and ultra-long-haul distances.
  • Operationalization of coherent pluggable optics into diverse environments like switches, routers, servers, and even 5G radios is something service providers want help solving. The management architectural framework for intelligent pluggables that the Open XR Forum is pursuing is a great example of vendors and service providers collaborating to reimagine network operations for a new generation of coherent optical engines.
  • Open optical and disaggregation are here to stay, with multiple sessions dedicated to open optical networking and disaggregation. It’s clear that service providers want more choice, faster innovation, and improved economics. And in a world with supply constraints, open also provides increased supply options for service providers.

Since not everyone in the industry could attend the event, I asked my Solutions Marketing teammates to write brief synopses on what they found most interesting or informative from this year’s conference.  I hope you find their unique perspectives illuminating.

Collage of Infinera photos from OFC 2022

High-performance Optical Evolution: Is Two Better Than One?

Paul Momtahan

A key topic at OFC was how far the industry could (and should) go in terms of baud rates and wavelength speeds. In terms of maximum baud rates, predictions ranged from around 160 Gbaud to 240 Gbaud. CMOS process nodes of either 3 nm or 2 nm should be capable of powering the digital ASIC/DSP at these speeds. In terms of the modulator, indium phosphide seems capable of evolving to these very high baud rates, though with some challenges related to the radio frequency (RF) interconnects. New modulator technologies, including thin-film lithium niobate modulators and plasmonic modulators, were also discussed as options. In terms of silicon photonics, while there was a lot of skepticism regarding its ability to evolve to these very high baud rates, it also had its backers and has surprised the industry before, so it cannot be counted out.  The key challenge for these ultra-high baud rates would appear to be the digital-to-analog converter (DAC) and the RF interconnects between the digital ASIC/DSP and analog electronics (drivers and transimpedance amplifiers).

One key question for the industry that was touched on in a number of presentations was whether it is better in terms of cost per bit and power consumption to have two lower-speed wavelengths or one higher-speed wavelength – for example, a single optical engine or pluggable that supports two 90 Gbaud 800 Gb/s wavelengths or one 180 Gbaud 1.6 Tb/s wavelength.  One interesting presentation compared two DACs with a DAC multiplexer and a single laser/modulator to two DACs and two lasers/modulators, asking if the DAC multiplexer would cost more than a second laser/modulator, but it did not provide an answer.

One techno-economic study compared single- and dual-wavelength approaches based on different assumptions regarding bits/s and baud rate cost scaling, with the single-wavelength approach winning if cost scaled with baud rate and the dual-wave approach winning if cost scaled with bits/s and the dual-wave cost penalty was not too high. Based on the sessions I attended, it looks like two 800 Gb/s wavelengths will be the first step toward 1.6 Tb/s, at least for pluggables. That said, in the case of Infinera’s ICE6 we already have a dual 800 Gb/s embedded optical engine that provides 1.6 Tb/s aggregate capacity.

Submarine Capacity: Please Sir May I Have Some More

Geoff Bennett

Scaling cable capacity is always a key topic in submarine networks, and this year’s OFC was no exception.  As we predicted in our blog last year, space-division multiplexing (SDM) is now the primary direction for cable development.  Fifth-generation coherent optical engines, such as Infinera’s ICE6, continue to exceed spectral capacity expectations, even on new SDM cables such as Dunant.  And the wet plant industry is confident in their roadmaps toward a future petabit-scale trans-Atlantic cable.

Looking beyond “multi-fiber” SDM, we saw interesting advances in multi-core fibers, and some interesting examples of multi-mode (or few-mode) transmission.  But it looks like multi-core fibers and multi-mode transmission may be beaten to the punch by the astonishing results published on hollow-core fibers (HCF), in particular by the U.K. company Lumenisity with their nested anti-resonant nodeless fiber (NANF).  To summarize, they are reporting attenuation that has now overtaken silica-based SMF-28 fiber.  And whereas silica fiber development has reached its limits, hollow-core fiber performance is still improving.  HCF is also interesting because the light within it propagates in air (a linear medium) rather than glass (a nonlinear medium).  Since nonlinear effects are the current limiting factor for signal-to-noise ratio in optical fiber, HCF could offer some real advantages for higher optical power levels.  In addition, HCF reduces latency on a given route by about one-third compared to conventional fiber.  Before you get too excited, NANF has only been manufactured in short lengths of a few kilometers so far, and it may take significant production developments to deliver thousands of kilometers of NANF at low enough prices.

So, the bottom line is that SDM is the preferred near-term option to scale submarine cable capacity.  Assuming hollow-core fiber can be made to work under the ocean (for example, resisting enormous water pressure), then we predict that this will be the most likely technology to come after SDM, rather than multi-core or multi-mode transmission.  But this is a rapidly evolving space, so we’re writing that second part of the prediction in dry-erase marker!

Open Networking: Can’t We All Just Get Along

Christian Uremovic

At this year’s OFC, there were several demonstrations highlighting open optical, multi-vendor, and networking interoperability demonstrations from industry-leading organizations. The Ethernet Alliance had demonstrations around 25 Gigabit Ethernet, 50 GbE, 100 GbE, 200 GbE, 400 GbE and 800 GbE interfaces and interconnected multiple booths across the show floor. OIF also showcased live multi-vendor interoperability demonstrations with 400ZR optics, co-packaging architectures, common electrical I/O (CEI) channels, Flexible Ethernet (FlexE) definitions, and Common Management Interface Specification (CMIS).

Open ROADM again demonstrated a live network and multi-vendor interoperability on the 400G transponder side and the optical layer side. Infinera provided our latest 400G technology in the GX G30 Compact Modular Platform, which directly interfaced Fujitsu and Cisco 400G transponder solutions. On the ROADM side, Infinera had the GX G30 ROADM directly interfacing Ciena and Fujitsu ROADM solutions. Path computation was performed by TransportPCE with GNPy to optimize performance, and through its RESTCONF interface the controller communicated with the University of Texas at Dallas Programmable Optical Network (PROnet) SDN orchestrator.

You can find out more in press releases from OFC and UT Dallas.

Network Software and Automation: Old Problems, Modern Approaches

Teresa Monteiro

Estimating quality of transmission (QoT) in optical networks continues to be a challenge. While traditionally QoT is estimated using detailed deterministic models and algorithms, this does not meet today’s network automation needs, because:

  • The models tend to be computationally intensive and therefore not suitable for the real-time operations required by a network controller.
  • The models typically focus on worst-case performance, ensuring path feasibility in real-world meshed networks through their lifetime, but preventing the dynamic and efficient use of resources imposed by current fast-paced traffic demands.
  • Most accurate models are vendor specific and do not apply well to open optical environments.

GNPy, a sponsored project of the Open Optical & Packet Transport (OOPT)/Physical Simulation Environment (PSE) working group of the Telecom Infra Project, is addressing the limitations outlined above through an open-source, community-developed software library for route planning and optimization. GNPy algorithms for optical fiber transmission are based on a Gaussian noise model and are fast, scalable, and accurate.

At OFC 2022, GNPy continued to show solid results, including an experimental validation of flexible transmission up to 800G using Nyquist subcarriers. But GNPy still has to address the detailed modeling of network equipment, in particular that of filtering penalties and crosstalk introduced by ROADMs, which have characteristics that vary significantly from device to device and from vendor to vendor.

A very different but alternative (or complementary?) approach is to use machine learning (ML) techniques for real-time, precise QoT estimation in open optical networks. This is very appealing, as it bypasses the need for detailed information on all network parameters.

Several presenters addressed this topic at OFC 2022, showing promising results and steady progress. But the discussion is still open on how to best gather large enough and representative datasets for real-world usage, while coping with data privacy issues.

One particularly interesting work presented at this year’s event discussed the application of explainable artificial intelligence (XAI) to QoT estimation. XAI aims to make ML decisions understandable by humans, and, with that, to improve transparency and increase trust in these types of systems.

The Two Worlds of Coherent Pluggables

Fady Masoud/Jon Baldry

The continued evolution of coherent pluggable optics was a common topic of discussion at this year’s OFC.  Seeing 400G and 800G press releases, presentations, and panel discussions, one thing became clear – there is not a one-size-fits-all approach to coherent pluggables.  In fact, there are at least two distinct coherent pluggable applications emerging. The first application addresses simple point-to-point fiber networks where the entire network is comprised of a single link of fiber between two points. This is a typical configuration for metro DCI applications. The second application is associated with metro and aggregation networks. These networks typically consist of numerous locations connected by a mesh of fiber and leverage ROADM technology.

To address these two applications, there are two distinct classes of pluggable optics emerging. While a very basic pluggable optical engine can support the first application, a more advanced optical engine is required for the second. To support mesh-based ROADM networks, optical engines require at least 0 dBm launch power to get through the add/drop structure of the ROADMs, low out-of-band noise to support a colorless/directionless add/drop structure, a higher optical signal-to-noise ratio to support amplifier chains, and signal shaping to support ROADM cascades.

Additionally, these network pluggables require a more sophisticated management paradigm to support the increased number of configurations, alarm correlation, and coordination with the optical line system. Most of these capabilities are not needed for simple point-to-point fiber networks. And contrary to what many think, the delineation between these two classes of optics is not just ZR and ZR+, as many ZR+ pluggables do not support the features critical for mesh-based ROADM network applications.

Creating a standard definition for network-grade pluggables and a more comprehensive approach to the management of coherent pluggables are two of the key objectives of the Open XR Forum.

Taking Control of Your Coherent Optical Engines

Jeff Babbitt

As my colleagues Fady and Jon mentioned, one key topic covered in multiple presentations was addressing the challenges that network operators face when controlling and managing coherent pluggable optics installed in third-party devices, such as routers, switches, or potentially even servers and 5G radios. Historically, coherent optics are part of the transport system, where control and management of coherent products is integrated into the overall management of the transport system. Traditional networks have a controller for the optical transport domain and a separate IP domain controller for the routers. So, what do we do with a coherent optical engine that now resides in the router domain (both physical and logically)? As more coherent pluggables are developed, including by Infinera, and this technology becomes more prevalent and adds advanced functionality, operationalization and management becomes an increasingly important topic.

Infinera’s Tim Doiron moderated a session on this topic in the Optica Executive Forum on Monday with panelists from Verizon, Windstream, Google, and Nokia.  In addition, the open network and streaming session at OFC presented multiple papers that explored the best way to manage this mixed-domain architecture. Several approaches were suggested, compared, and contrasted. One promising approach is that of extending the scope of transport management to include the control of pluggables installed in the router.

Infinera is collaborating with members of the Open XR Forum in rethinking the management of intelligent coherent pluggables.  The emerging management architectural framework takes a holistic approach and strives to create independence from the network infrastructure that the intelligent pluggable is installed in or the line system it runs over. In alignment with this ongoing process, Infinera also announced its Open Optical Toolkit – a set of lightweight, cloud-native applications that simplify deployment and management in any network element and over any line system.

Earthquake Detection: Feel the Earth Move Under Your Cable

Geoff Bennett

A hot topic at OFC this year was the idea of using existing optical cables as seismic sensors – on land but also, perhaps more importantly, under the sea.  The key point is that undersea earthquakes cause tsunamis that can devastate coastlines and kill thousands of people.  The more warning time we have, the more people can be saved, and literally every second counts.  Every 200 km closer the sensor is to the epicenter of the quake means an additional minute of warning.  And we must be able to use the hundreds of existing subsea cables and repeaters around the world today, rather than wait for new types of sensors to be installed.  The session on network intelligence included fascinating examples of terrestrial seismic sensing by Mark Englund of FiberSense and Glenn Wellbrock of Verizon, as well as subsea examples by Eric Breverman at Google and Pierre Mertz of Infinera, who described an incredibly successful trial using ICE6 on the 10,500-km Curie subsea cable between Los Angeles, CA and Valparaiso, Chile.

The bottom line here is that terrestrial sensing can be used to detect and act against unauthorized digging close to existing cables.  And existing submarine networks could provide a way to deliver timely warnings of submarine earthquakes to coastal communities around the world.

That’s a Wrap

As I headed to the airport on Friday morning, I felt both exhaustion and a sense of pride.  The last two years have been difficult and disruptive to so many families and their loved ones.  But throughout it all our industry has worked tirelessly to deliver more capacity and enhanced networks to connect people and applications to the cloud and each other – every single day.  Finally getting to meet in person with teammates, customers, partners, analysts, and industry colleagues was just what many of us needed. For all those in attendance, I hope you felt the same.  If you couldn’t attend, I hope you found this blog helpful – and my team and I look forward to seeing you at another in-person event in 2022.

Safe travels everyone!