6 Networking Trends You Might Have Missed at OFC Last Week
June 17, 2021
By Tim Doiron (and Solutions Marketing Team)
Sr. Director, Solution Marketing
OFC 2021 is officially in the books. While I really missed seeing colleagues and friends across the industry face to face, the virtual format was a strong alternative and certainly provided an opportunity for me and the rest of Infinera’s Solutions Marketing team to attend far more sessions and presentations than we could have done live from the show floor.
I moderated an Edge panel the week before OFC as part of the OIDA Executive Forum as well as the highly anticipated OFC Market Watch: State of the Industry Panel. The OFC panel consisted of two financial analysts and three industry analysts: Alex Henderson, Needham; George Notter, Jefferies; Vlad Kozlov, LightCounting; Andrew Schmitt, Cignal Ai; and Jimmy Yu, Dell’Oro.
While the industry faces some challenges like component production ramps and global supply chain evolution, the market for optical transmission equipment is strong and the industry is healthy. With commercial adoption of fifth-generation 800G coherent DWDM embedded optical engines ramping; new pluggable coherent modules emerging, including point-to-multipoint XR optics; and evolutions in software intelligence analysis and automation in areas like machine learning to maximize quality of transmission, I can’t remember a more exciting time to be in this industry.
Since time is short and not everyone could attend the virtual 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. See you in sunny San Diego for OFC 2022.
Flying Close to the Shannon Limit
A number of talks focused on where we are in relation to the Shannon limit and what the options are for getting closer to it. Today’s high-end coherent transceivers are around 2 dB away from the Shannon limit, with about 1 dB of this gap coming from the DSP. Advances in the CMOS process node, from 7 nm to 5 nm, 3 nm, and beyond, will enable better modulation and forward error correction in the DSP, which will help to close this gap, though at a cost of increased power consumption relative to the same CMOS process node without these improvements.
Reducing the implementation penalty with the ADC/DAC, RF interconnects, and other components will also help. An alternative option for increasing capacity is to reduce the margin based on careful monitoring of the margin and adjustments to the transponder baud rate and modulation. Another option for increased capacity is monitoring for crosstalk and filtering penalties and then adjusting the baud rate to match that actual ROADM cascade passband. A final option presented was based on reducing the wavelength roll-off from ~6%-15% today down to 1%, making the spectral shape of the wavelengths squarer so that they can be packed closer together.
Fiber Capacity Was All the Rage at OFC
Many of the talks this year looked at where we go next when today’s fibers are “full.” C+L transmission is already used commercially, but transmission in other bands (O-, E-, S-, and U-bands) will be limited by higher attenuation and the lack of commercial amplification options. Going parallel is an obvious option, and for network operators whose ducts have limited capacity, we heard about reduced-cladding-diameter fibers, where a 20% reduction in fiber diameter can increase the number of fiber pairs by 50%. Beyond this is the fabrication of multicore fiber.
Uncoupled multicore fibers are like deploying multiple, parallel fibers in a single strand, but these cores have to be amplified independently. Closely coupled multicore fibers could allow multiple cores to use the same fiber amplifier because the information interacts across the cores, but then we have to reverse this mixing of information using a MIMO and DSP technology.
Finally, we are seeing hollow-core fiber (HCF) being deployed in financial networks that require ultra-low latency. Attenuation in HCF is very high today – around 2 dB/km compared to about 0.2 dB/km for SMF – so reach is limited. Multicore and HCF will require a completely new fiber ecosystem of connectors, installation training, and splicing technologies, so it may be some time before we see widespread use of these fiber types.
Riding the Innovation Wave with the Open XR Forum
The foundation of the Open XR Forum is considered by many to be “the” announcement of OFC 2021. It marks a significant milestone in XR optics’ journey since it was first introduced to the market back in 2019. This is an industry collaboration that will accelerate the widespread adoption and deployment of XR optics’ innovative technology by a broad range of network operators in a wide variety of applications.
Open XR Forum members will ensure the products and services developed align with existing standards and operational models by driving standardization of network interfaces and electro-mechanical hardware interfaces, demonstrating interoperability, establishing multi-sourcing potential, and developing and publishing new specifications as needed.
The foundation of the Open XR Forum further fuels XR optics’ great market momentum. This was witnessed numerous times during OFC 2021 sessions, such as when Andrew Lord, Head of Optical Networks Research at Applied Research, BT, highly talked about XR optics’ innovation and its predicted positive impact in his Future Optical Networks in a 10 Year Time Frame (M2A.3) session.
It’s great to see our customers and prospects talk about our technology, but you know it’s groundbreaking when other component manufacturers or even competitors include slides on XR optics in their OFC presentations, which was the case for Fujitsu Optical Components and Futurewei (owned by Huawei). XR optics is rocking the optical industry, and more great things are still to come.
Pushing the Boundaries of Pluggable Optics in Metro
One significant topic at OFC this year was the progress the industry has made in the use of coherent 400G pluggable optics in a broad range of metro networking applications. With 400G ZR optics starting to ramp in short-reach, high-capacity DCI applications and 400G ZR+ optics starting to be deployed in metro aggregation networks, there was considerable discussion around where the industry goes next:
1 – Form factor and performance: While QSFP-DD formats match router ports and other devices, CFP2-DCO formats continue to provide the better performance needed for metro core and regional applications, with CFP2-DCO ZR+ 1500-km reach being announced at the show.
2 – 800G ZR/ZR+: Numerous presentations discussed the next step to 800G in both ZR and ZR+ applications. Once ports in high-capacity switches and routers jump to 800G, we’ll see an urgent demand for these devices. OSFP seems to be an early leader in the race for the preferred format for 800G pluggables, although others such as QSFP-DD800 are also in the race. Predictions varied on availability of these optics, and the complexity of developing them should not be underestimated.
3 – Point-to-multipoint optics: It was interesting to see point-to-multipoint optics covered by many presenters, not just those from Infinera. These optics promise to play a critical role in shaping the future of high-speed metro networks. My colleague Fady Masoud has also covered this topic above.
AT&T and University of Texas Open Networking Demo
Open networking is further evolving, and the University of Texas at Dallas, together with AT&T Labs and Orange Labs, invited open optical networking vendors to participate in an Open ROADM demonstration at OFC 2021. The Open ROADM MSA standard is defined for both the optical data plane and the control plane. The key objective of the demo was to demonstrate and test wavelength interworking as well as optical layer interworking between vendors. Participants included Ciena, Cisco, Fujitsu, Infinera, Juniper, Nokia, and Ribbon, together with the researchers at AT&T, Orange, and the OpNeAR laboratory at the University of Texas at Dallas.
The whole setup was managed by open-source software, such as Openstack and OpenDaylight TransportPCE controller, among others. Through its northbound RESTCONF interface, the TransportPCE controller communicated with the UT Dallas PROnet (Programmable Optical Network) SDN orchestrator, providing a single point of control and coordination of resources across the Open ROADM optical layer and the client layer.
This standardized and well-defined interoperability demonstration showcased the ability to help network operators reduce operational efforts to ease new technology introduction and integration. Infinera is committed to open networking, and we are pleased to drive these requirements with our Open ROADM membership and welcome the opportunity to showcase our market-leading technology. More information can be found in the OFC paper and the video below.
Machine Learning Looks Promising
I was glad to see CSPs and ICPs, from BT to AT&T, Telia Company, euNetworks, and Google alike, highlighting a surge in network automation. Networking needs unveiled by the pandemic, such as faster response to demands and remote capacity turn-up, acted as catalysts for automation. And automation is here to stay: it’s the way to deal with ever more complex networks and to keep up with customer expectations and service-level objectives.
Looking ahead toward autonomous networking, I noticed the large number of sessions on machine learning (ML) at OFC. ML applications in optical networking are somewhat “late to the ML party,” but OFC showcased a lot of promising work, in particular on the use of these methodologies to predict quality of transmission.
Some issues are holding back the productization of these applications, namely:
- Data quality and availability, including concerns on data sharing and protection
- Identification of adequate validation KPIs and assessment of the algorithms’ precision
- Definition of priority use cases: where best to start and how to build a business case
- User trust in ML solutions
But as optical networking benefits from lessons learned in fields where mature ML solutions exist, I am optimistic we will see an uptake in field deployments in the next three to five years.
That’s a Wrap
One thing was clear from many of the sessions, the trend of open networking continues as network operators want more choice at almost every layer, especially between optical transponders and line systems. These advancements in open will yield faster innovations and an ecosystem of solutions that best serve the customers needs. See you next year, in person!