The Optical Networking and Communication
Conference & Exhibition

San Diego Convention Center,
San Diego, California, USA

Data Center Presentations and Events

Programs below with an asterisk are on the show floor. Register for Exhibits Pass Plus and visit the show floor for FREE.

ata Center Schedule Sunday,
8 March
Monday,
9 March
Tuesday,
10 March
Wednesday,
11 March
Thursday,
12 March
Workshop: Are Radical Photonic Devices and Architectures Needed for Future Data Centers? * 16:00 - 18:30        
Workshop: Does Disaggregation Support Data Center Evolution?* 16:00 - 18:30        
Edge Computing (M1A)   08:00 - 10:00      
Chip-to-chip Optical Interconnects (M1H)   08:00 - 10:00      
Ethernet Interoperability and Deployments – New and Legacy Solutions Work Together*   10:15 - 11:15      

High Order Direct Detect Formats (M4F)

  16:30 - 18:15       
Data Center Summit*     11:30 - 13:45    
400ZR Specification Update*     13:30 - 14:30    

Intra Data Center Networks I (T3K)

    14:00 - 16:00    
Embedded Optics and How They Should be Implemented to Support the OEM Eco-system*     15:00 - 17:00    
Rump Session: When Will Co-packaged Optics Replace Pluggable Modules in the Datacenter?     20:30 - 21:30    

Short-reach Interconnects (W1D)

      08:00 - 10:00  

Intra Data Center Networks II (W1F)

      08:00 - 10:00  

Datacentre Infrastructure and Metrology (W3G)

      14:00-16:00  
New High-bandwidth, Non-DSP Interface for Data Center and Campus Interconnects*       15:00 - 16:00  
Market Watch Panel IV: What Is Next for Data Center Interconnects?*       15:30 - 17:00  
Market Watch Panel V: Inside the Data Center*         10:30 -12:00
Beyond 400ZR….What Comes Next?*         11:00 - 12:00
System Evaluation of On-board Optics*         11:30 - 12:30

Optimizing Coherent Transponders (Th3E)

        14:00- 16:00

Organizer:

Maura Raburn, Google, USA; Kenya Suzuki, NTT Device Innovation Center, Japan; Yikai Su, Shanghai Jiao Tong University, China

Description:

What is in store for future datacenters? 
(When) will optical circuit and packet switching dominate?  Or will more traditional architectures and devices meet future speed, cost, scale-ability, and latency requirements? 
Experts with opposing perspectives will debate our future.

Speakers:

  • Architectures:
    • Ken-Ichi Sato; Nagoya University, Japan
    • George Porter; University of California San Diego, USA
    • Lena Wosinska; Kungliga Tekniska Hogskolan, Sweden
  • DC Operator:
    • Chongjin Xie; Alibaba Group, USA
    • Francesca Parmigiani; Microsoft Research Ltd, UK
  • Devices:
    • Shifu Yuan; Calient Technologies Inc., USA
    • Salah Ibrahim; NTT Photonics Laboratories, Japan
    • Mohan Kalkunte; Broadcom Ltd., USA

Organizer:

Michela Svaluto Moreolo, Centre Tecnològic Telecomunicacions Catalunya, Spain; Madeleine Glick, Columbia University, USA; Ken-ichi Kitayama, The Graduate School For The Creation Of New Photonics Industies, Japan

Description:

The concept of disaggregation is increasingly popular in both Datacom and Telecom, driven by the ever increasing capacity demand at reduced cost, as a promising candidate towards a more efficient resource utilization, improved flexibility, scalability and programmability. This workshop aims at exploring opportunities and enablers answering questions related to the disaggregation paradigm for supporting data center (DC) evolution from different perspectives, involving its impact on architectural, networking and management aspects as well as technological ones. Could this new paradigm achieve/provide the promised opportunities? Which are the enabling technologies? Are they mature enough? Which are the most promising for an actual increase of efficiency, cost/power saving, flexibility? What are the underlying challenges? What are the drawbacks? Could disaggregation be the enabler for converged inter/intra data center optical networks?

Specifically, the workshop will explore if an appropriate network architecture could be a key enabler of DC disaggregation. How is the inevitable additional latency of the disaggregated network compensated/mitigated/ alleviated? How is the bandwidth requirement addressed?

What are the architectural, control and management ramifications of disaggregation? How can these be addressed? Is there resistance from the user (data center, telecom operators) community? Which is/are the best architectural choice(s) in support of DC disaggregation?

How can scalability be efficiently addressed? Where is the bottleneck? Are the solutions being proposed scalable or a temporary stop gap? What are the ramifications of disaggregation with respect to energy efficiency? Is it an energy cost or an energy saver? Is disaggregation viable now and if not, what is the block? Is it just not yet?

The technology implications at the system, subsystem and infrastructure level represent another key point towards the data center evolution in the framework of disaggregation paradigm. What is the role of the switching infrastructure? How should it be designed to support and ease disaggregation and what are the challenges? Are new hardware and specialized components needed? What is the role of photonics? What is the vendor perspective? Are there alternative non disaggregated solutions that meet cost/performance targets? Are there critical applications that would suffer performance degradation with disaggregation?

Speakers:

  • Victor Lopez; Telefonica, Spain
    Enabling an Open Network Ecosystem: SDN and Whiteboxes
  • Hitesh Ballani; Microsoft, USA
  • Ling Liao; Intel Corporation, USA
    SiPh Based Co-package Optics for Data Center Disaggregation
  • Shu Namiki; National Institute of Advanced Industrial Science and Technology (AIST), Japan
    Disaggregation and Automation of Optical Layer Switching for Converged Compute and Network
  • Nicola Calabretta; Eindhoven University of Technology (TUE), Netherlands
    Disaggregated Data Center Architecture Enabled by Nanoseconds Controlled Photonic Switches
  • Georgios Zervas; University College London (UCL), UK
    MONet:  Memory over Optical Network at Cluster Scale - From Physical Layer to Application Performance
  • John Shalf; Lawrence Berkeley National Laboratory, USA
    Diverse Accelerators Are Coming: What Are the Alternatives for Integrating them into the Datacenter?

Telemetry-Driven Optical 5G Serverless Architecture for Latency-Sensitive Edge Computing (M1A.1)
08:00 AM - 08:15 AM
Latency-sensitive serverless subfunctions are optimally deployed at edge and cloud according to telemetry-retrieved data from the 5G transport infrastructure. Once deployed, serverless functions provided extremely fast invocation time of less than 450ms.
Authors: I. Pelle, B. Sonkoly, MTA-BME Network Softwarization Research Group, Budapest, HUNGARY|F. Cugini, CNIT, Pisa, ITALY|F. Paolucci, Scuola Superiore Sant’Anna, Pisa, ITALY

Flexible Optical Network Enabled Hybrid Recovery for Edge Network with Reinforcement Learning (M1A.2)
08:15 AM - 08:30 AM
The proposed hybrid recovery utilizes flexible optical network with reinforcement learning to recover IP fault for edge network. The testbed experiments indicate, the recovery time is 20% of rerouting-based strategy for a heavy-loaded network.
Authors: M. Lian, R. Gu, Y. Qu, Z. Wang, Y. Ji, Beijing Laboratory of Advanced Information Network, Beijing University of Posts and Telecommunications, Beijing, CHINA

Multi-layer Network Slicing for Accelerating Business Velocity for Edge Computing (M1A.3) - Invited
08:30 AM - 09:00 AM
Abstract not available.
Authors: A. Nakao, Interfaculty Initiative in Information Studies, The University of Tokyo, Tokyo, JAPAN

Deep Reinforced Energy Efficient Traffic Grooming in Fog-Cloud Elastic Optical Networks (M1A.4)
09:00 AM - 09:15 AM
We propose a novel energy efficient traffic grooming algorithm based on deep reinforcement learning in fog-cloud elastic optical networks. Simulation results show that it can achieve much lower energy consumption than the state-of-art algorithm.
Authors: R. Zhu, S. Li, P. Wang, L. Li, A. Samuel, Zhengzhou University, Zhengzhou, HENAN, CHINA|Y. Zhao, Beijing University of Posts and Telecommunications, Beijing, CHINA

Multi-stage Aggregation and Lightpath Provisioning of Geo-distributed Data over EON Assisted by MEC (M1A.5)
09:15 AM - 09:30 AM
A multi-stage aggregation and lightpath provisioning algorithm is proposed for geo-distributed data in EON assisted by MEC. Simulation results show the algorithm can reduce the job completion time and bandwidth consumption.
Authors: Z. Liu, J. Zhang, Z. Guo, Y. Ji, Beijing University of Posts and Telecomm, Beijing, BEIJING, CHINA

Remote Human-to-Machine Distance Emulation through AI-Enhanced Servers for Tactile Internet Applications (M1A.6)
09:30 AM - 09:45 AM
We alleviate the master-slave distance limitation of human-to-machine applications by forecasting and pre-empting haptic feedback transmission. Results show 99% accuracy in detecting touch events and 96% accuracy in forecasting feedback from different slave materials.
Authors: S. Mondal, L. Ruan, E. Wong, University of Melbourne, Melbourne, Victoria, AUSTRALIA

Demonstration of Geo-Distributed Data Processing and Aggregation in MEC-empowered Metro Optical Networks (M1A.7)
09:45 AM - 10:00 AM
We experimentally demonstrate a geo-distributed data processing and aggregation (GDPA) scheme in the MEC-empowered metro optical networks. The demonstration results show that the proposed scheme can improve resource utilization and reduce average job completion time.
Authors: J. Zhang, L. Cui, Z. Liu, Y. Ji, Beijing Univ of Posts & Telecom, Beijing, BEIJING, CHINA


Co-packaged TeraPHY Optical I/O Enables Next Generation of Data Center Applications (M1H.1) - Invited
08:00 AM - 08:30 AM
Abstract not available.
Authors: V. Stojanovic, Ayar Labs, Emeryville, California, UNITED STATES

Phase Noise Spectral Properties Across Individual Comb Lines in Quantum-Dot Mode-Locked Lasers (M1H.2)
08:30 AM - 08:45 AM
We study phase-noise spectral properties of comb lines from a QD-MLL, show that their large linewidth variability attributes to the low-frequency phase variations, and has minimal effect on coherent system performance at practical symbol rates.
Authors: M.A. Al-Qadi, R. Hui, University of Kansas, Lawrence, Kansas, UNITED STATES|M. O'Sullivan, R&D, Ciena Corporation, Ottawa, Ontario, CANADA|C. Xie, R&D, Alibaba Group, Sunnyvale, California, UNITED STATES

Experimental Demonstration of PAM-4 Transmission through Microring Silicon Photonic Clos Switch Fabric (M1H.3)
08:45 AM - 09:00 AM
We present the first experimental demonstration of a 25 Gbps optical PAM4 signal transmission through a microring-based Clos topology under realistic operating conditions. We observe a 1.1-dBm power penalty at the bit error rate of 1.03×10−7
Authors: L. Dai, Y. Hung, Q. Cheng, K. Bergman, Lightwave Research Laboratory, Harrison, New Jersey, UNITED STATES


Session Description

This panel will discuss the testing that results in successful end user deployments. It will also address the evolution of new PAM4 based solutions – 50GbE and above – entering the market and the advancements supporting Ethernet interoperability testing for the successful deployment and integration into legacy networking environments. From switch, server, and storage arrays attached over the multitude of interconnect options, to the advanced storage applications, seamless integration and interoperability for all participants in the ecosystem is a critical “must have” for meeting product deployment schedules, reducing post-sales support issues, and ensuring customer expectations are exceeded. The panel will review the challenges of testing the myriad of Ethernet speeds, and the impact of emerging, mission critical applications utilizing legacy and new, 50GbE and higher speed Ethernet networks.

Moderator David RodgersConference Chair, Teledyne LeCroy PSG - Sr. Product Marketing Manager, Ethernet Alliance, USA

Speakers

  • Paul BrooksDirector  - Lab & Production Strategy, Viavi Solutions, Germany
  • Sheng HuangDirector of Applications Engineering, Credo
  • Shawn NichollDirector of Design Engineering, Xilinx
  • Matt TraversoPrincipal Engineer, Cisco Systems, USA

280 Gb/s IM/DD PS-PAM-8 Transmission Over 10 km SSMF at O-band For Optical Interconnects (T3K.1)
04:30 PM - 04:45 PM
We experimentally demonstrated single-lane 200G+ IM/DD PAM-N system at O-band using SOA and probabilistic shaping (PS) for high-speed short reach optical interconnects. 280 Gb/s PS-PAM-8 signals can transmit over 10 km SSMF.
Authors: J. Zhang, K. Wang, Y. Wei, L. Zhao, W. Zhou, J. Xiao, J. Yu, Fudan University, Shanghai, CHINA|B. Liu, X. Xin, Beijing University of Posts and Telecommunications, Beijing, CHINA|F. Zhao, Xi’an University of Posts and Telecommunications, Xi'an, CHINA|Z. Dong, Huaqiao University, Xiamen, CHINA

30 Gbaud 128 QAM SSB Direct Detection Transmission over 80 km with Clipped Iterative SSBI Cancellation (T3K.2)
04:45 PM - 05:00 PM
We demonstrate a novel SSBI cancellation technique operable without digital upsampling for a 30 Gbaud 128 QAM SSB transmission with a record low CSPR of 5 dB, showing 4.6 dB performance improvement compared to the Kramers-Kronig scheme
Authors: S.T. Le, V. Aref, K. Schuh, Nokia Bell Labs, Aberdeen, New Jersey, UNITED STATES|H. Nguyen Tan, Da Nang University, Da Nang, VIET NAM

Novel Optical Field Reconstruction for IM/DD with Receiver Bandwidth Well Below Full Optical Signal Bandwidth (T3K.3)
05:00 PM - 05:15 PM
We propose a novel signal reception scheme for IM/DD enabling optical field reconstruction. We experimentally demonstrate 60-GBd PAM-4 transmission over 80-km without active and passive optical managements, with 33-GHz electrical bandwidth at transmitter and receiver.
Authors: Q. Hu, R. Borkowski, M. Chagnon, K. Schuh, F. Buchali, H. Bülow, Nokia Bell Labs, Stuttgart, GERMANY

Demonstration of 214Gbps per lane IM/DD PAM-4 transmission using O-band 35GHz-class EML with advanced MLSE and KP4-FEC (T3K.4)
05:15 PM - 05:30 PM
A single-wavelength single-polarization 35GHz-class (112Gbps-class) commercial EML-based IM/DD 214Gbps PAM4 signal transmission is experimentally demonstrated. By using advanced MLSE with low complexity and power consumption, the BER is below standard KP4-FEC requirement of 2×10-4.
Authors: W. Wang, Z. Huang, B. Pan, H. Li, G. Li, J. Tang, Y. Lu, Huawei Technologies Co. Ltd., Beijing, CHINA

160-Gb/s Nyquist PAM-4 Transmission with GeSi-EAM Using Artificial Neural Network Based Nonlinear Equalization (T3K.5)
05:30 PM - 05:45 PM
We experimentally demonstrate optical interconnects of PAM-4 signal with a single lane bit rate of 160Gb/s generated by a compact silicon based GeSi electro-absorption modulator using artificial neural network based nonlinear equalization.
Authors: L. Zhang, F. Yang, H. Ming, X. Ruan, Y. Li, F. Zhang, Peking University, Beijing, CHINA|Y. Zhu, ZTE, Shanghai, CHINA

Why Data Science and Machine Learning Need Silicon Photonics (T3K.6) - Invited
05:45 PM - 06:15 PM
Training deep neural networks demands vast amounts of computation, provided by large distributed systems. The increasing demand for bandwidth will exceed the limits of electrical and non-integrated optical signaling and will require integrated optics in the future.
Authors: B. Klenk, L. Dennison, NVIDIA Corporation, Mendon, Massachusetts, UNITED STATES


Keynote: Reducing Power in Network Infrastructures

11:30-12:15

Moderator: Karen I. MatthewsTechnology and Market Development Manager, Corning Research & Development Corporation, USA

Description

Increasingly, the electrical power consumed by network infrastructure is negatively impacting the efficiency of large scale datacenter designs.  Over the last decades, network power, as a percentage of overall DC power, has begun to rival the power consumption of cooling and other support systems.  In this presentation we will examine the contributing causes of the growing network infrastructure power trend and investigate some of the proposed solutions to address the issue. 

Jeffrey L. Cox, Partner Director Network Architecture, Microsoft Corporation, USA

Panel: Data Center 2020 – Less Hyper-scale and More Co-location and Compute at the Edge?

12:15-13:45

Moderator: Robert Blum, Director of Marketing and New Business Silicon Photonics Product Division, Intel, USA

Description

A lot of focus has been put on the economy and scale of hyper-scale data centers. The last decade witnessed a massive scale out of data centers, some of which contain upwards of a hundred thousand severs and require more than a hundred Megawatts of power. However, security, privacy and data sovereignty considerations have also required the use of much smaller data centers closer to the end user. And with the roll out of 5G and the need for more and more data tied to time critical applications such as real time maps for autonomous driving, edge and co-location data centers are becoming an increasingly critical component of the network as well. 

This panel will discuss the latest trends in data centers from an infrastructure and networking hardware perspective. What vectors and opportunities exist to reduce power consumption, size, and cost, what architectures are being considered inside data centers, and how are data centers evolving and being linked together at the regional and global scale to address the needs of the new decade?  


Speakers

  • Ed Baichtal, Solutions Architect, Equinix, USA
  • Dan PittVice President, MEF Forum, USA
    SD-WAN: Genesis of Edge Computing Platforms

    We all agree that Edge Computing will open new vistas for applications, services, and revenues. Cool designs have been proposed for disaggregated computing platforms and software structures for VNFs and CNFs from all sorts of authors. But what will get these computing platforms out in all the network edge locations where they are desired? We think the answer is SD-WAN, as that is the commercial service that even now is in rapid deployment. Between Provider Edge and Customer Edge, SD-WAN will put computing power in enough places to drive the growth of these promising and highly diverse new applications and services. No other single application or service will have the critical mass. In this very brief talk we also describe how SD-WAN standards are enabling and accelerating the deployment of these very platforms.
  • Saurabh SandhirVice President, Product Management, Nuage/Nokia, USA
    Distributed Telco Cloud: Drivers and Enablers

    The talk will describe the emerging edge computing landscape from a telco service provider point of view, look at the nature of the distributed cloud emerging as a result and talk of underlying network infrastructure needed for it. We will look at a couple of customer examples from the Nuage journey to highlight the use cases.
  • Zuowei ShenSenior Staff Hardware Engineer, Google, USA
    Latest Trends and Opportunities in Hyperscale Data Center
    We will review technology evolution of optical/electrical components and packaging needed to further scale performance and efficiency in data centers. 
  • Chongjin XieSenior Director, Alibaba, USA
    The Impact of Edge Computing on the Architecture of Data Centers

    The development of edge computing may have a significant impact on data center architecture, including the size and networks of edge and main data centers.

Session organized by 400ZR Specification Update (OIF)

Session Description

A panel of industry experts representing the coherent eco-system will discuss and debate the conflicting demands for a near-term, high-volume, interoperable, moderate reach, coherent 400G optical link. The status of the OIF’s project to define a 400ZR link specification will also be provided. A panel discussion including representatives of the DSP, optics, equipment and end user communities will ensure this will be a lively discussion. Participate in a conversation with the OIF on the vital need for an interoperable 400G Coherent link and gain an update on the project status.

Moderator:

Karl GassOIF PLL WG Vice Chair Optical, OIF, USA


Demonstrating Optically Interconnected Remote Serial and Parallel Memory in Disaggregated Data Centers (T3K.1)
02:00 PM - 02:15 PM
Remote serial and parallel memory using memory-over-network bridge and optical switched interconnect is demonstrated. Remote memory bandwidth of 93% (HMC) and 66% (DDR4) of the local 3.2 and 3.7 GB/s bandwidth is showcased.
Authors: V. Mishra, J.L. Benjamin, G.S. Zervas, University College London, London, UNITED KINGDOM

Analysis of Service Blocking Reduction Strategies in Capacity-limited Disaggregated Datacenters (T3K.2)
02:15 PM - 02:30 PM
Disaggregated DCs offer multiple benefits. However, transmission capacity limitations at blade level can severely degrade their performance. We analyze several strategies to enhance their service acceptance.
Authors: A. Pagès, F. Agraz, S. Spadaro, Universitat Politècnica de Catalunya (UPC), Barcelona, SPAIN

Advanced Software Architectures and Technologies in High Performance Computing and Data Centers (T3K.3) - Invited
02:30 PM - 03:00 PM
This paper reviews advanced software architectures and technologies that support innetworking computing and improve the overall performance of data centers and high-performance computing clusters; the ability to converge software and hardware allows for new solutions, such as artificial intelligence, to be deployed massively.
Authors: J. Vegas Olmos, L. Liss, T. Oved, Z. Binshtock, D. Goldenberg, Mellanox Technologies, Roskilde, DENMARK

Real-Time Node Local Control for Ultra-Dynamic and Deterministic All-Optical Intra Data Center Networks (T3K.4)
03:00 PM - 03:15 PM
We enable ultra-dynamic features in scheduled optical data centers through a novel control mechanism local to each node. We experimentally show sub-μs resource allocation, at least halving distributed computing application completion time.
Authors: M. Szczerban, J. Estarán Tolosa, N.D. Benzaoui, H. Mardoyan, Y. Pointurier, Nokia Bell Labs, Nozay, FRANCE

Coherently Sub-Grouped µDC-Pod and -Interconnect with Analogue EML Transceivers Operated in TDMA (T3K.5)
03:15 PM - 03:30 PM
We exploit an IM/DD transmitter as coherent receiver for filterless micro-datacenter pods and their interconnect. A transistor-outline EML performs coherent homodyne reception under a 240kHz TDMA frame with 139ns guard interval between free-running transmitters.
Authors: B. Schrenk, N. Vokic, D. Milovancev, AIT Austrian Institute of Technology, Vienna, AUSTRIA|P. Bakopoulos, Mellanox Technologies Ltd, Yokneam, ISRAEL|F. Karinou, Microsoft Research Ltd., Cambridge, UNITED KINGDOM

Data Analytics Practice for Reliability Management of Optical Transceivers in Hyperscale Data Centers (T3K.6)
03:30 PM - 03:45 PM
There are limitations when directly interpreting reliability information of optical transceivers from manufacturers to end users. Data analytics in a large optical transceivers’ population is studied for data center operators with a case study.
Authors: J. Li, Alibaba Group, Bellevue, Washington, UNITED STATES|Z. Wang, C. Wang, Q. Chen, P. Wang, R. Lu, Alibaba Group, Hangzhou, Zhejiang, CHINA|C. Wang, S. Fu, Huazhong University of Science and Technology, Wuhan, Hubei, CHINA|C. Xie, Alibaba Group, Sunnyvale, California, UNITED STATES

Scaling HPC Networks with Co-packaged Optics (T3K.7)
03:45 PM - 04:00 PM
We propose an HPC network architecture with co-packaged optics enabling 128-port 51.2-Tb/s switches. Simulations for a >34,000-accelerator system show up to 11.2x throughput improvement over the Summit supercomputer, opening the way to direct-network-attached GPUs.
Authors: P. Maniotis, L. Schares, B. Lee, M. Taubenblatt, D. Kuchta, IBM TJ Watson Research Center, Yorktown Heights, New York, UNITED STATES


Session Description

With the next generation of Ethernet switch chips from vendors like Broadcom, Huawei and Avago, the need for better integration to support the interface bandwidth and density of I/O becomes apparent. Embedded optics standards from COBO and CPO are compared to co-packaged optics from Ayar Labs.  All are competing for mindshare and the hearts of switch designers. Optical transceiver vendors also have a stake in this market with planned, higher levels of integration in a pluggable format. This panel will allow all of the players to state their case and we will also hear perspectives of some users and OEMs. 

Moderator

Steve PloteOptics Consulting Engineer, Nokia, Inc., USA

Panelists

  • Brad BoothPresident, COBO; Network Hardware Manager, Microsoft Azure, USA
    The Death of Faceplate Optics

    There is a lot of hype around moving optics away from the faceplate of switches and networking equipment, but is that really necessary? This presentation focuses on why this trend is inevitable and how the optics and networking industry needs to embrace this change.
  • Ron Johnson, Senior Director of Architecture and Product Management, Cisco Systems, USA
    Commercial Implications of Embedded Optics

    This talk will explore the commercial implications of transitioning from pluggable to embedded optics, how this impacts your node design and things you should consider.
  • Katharine Schmidtke, Co-packaged Optics Collaboration (CPO) Rep and Facebook, USA
    Co-packaged Optics Ecosystem for Hyperscale Data Center Applications
  • Vladimir StojanovicChief Architect and Co-Founder, Ayar LabsUSA
    TeraPHY Optical I/O - The New Universal Chip to Chip I/O Solution
    Optics has reached the potential to provide the rack and row-scale connectivity at the energy and latency cost of electrical in-package interconnects.  To achieve this, photonics has to be crafted to fit the electronics SoC packaging ecosystem and embrace the inherent bandwidth density and energy-efficiency advantages by utilizing monolithic integration with CMOS and advanced photonic structures. Here, we'll describe how Ayar Labs' TeraPHY Optical I/O technology is positioned to take the mantle from electrical SerDes as the next universal chip to chip I/O solution.
  • Martin ZirngiblChief Technologist, II-VIUSA
    What Does Co-packaging Mean for the Datacom Ecosystem?
    Co-packaging will fundamentally alter today’s supply chain in the datacom industry, both from a technology as well as business aspect. We will discuss some of the challenges as well as opportunities in a world where switches no longer use pluggable transceiver but where the optics is directly hardwired to the switch ASIC.

Organizers: Chris Cole,Luminous Computing, USA; Dan Kuchta, IBM Research, USA

Provocateurs:

  • Joris Van Campenhout; IMEC, Belgium
  • Peter De Dobbelaere; Cisco/Luxtera, USA
  • Jane Gu; University of California Davis, USA
  • Shu Namiki; AIST, Japan
  • Zuowei Shen; Google, USA
  • James Stewart; Facebook, USA
  • Rob Stone; Broadcom, USA
  • Greg Walz; Molex, USA
  • Zhiping Yao; Alibaba, China
Description: 

 

A major limitation of today’s switch cards is the PCB electrical connection between the ASIC and front panel pluggable optics. As Baud rate increases the PCB link SerDes power increases to overcome frequency dependent copper trace and connector losses. Further, ASIC bandwidth is constrained by physical limitations of the MCM ball grid array and pluggable module electrical connector I/O count. Co-packing the ASIC and optics on a common MCM promises to remove this limitation. Over the past decade, industry has been working to make this approach real, with the pluggable paradigm holding firm. With each new generation of switch ASICs, the demise of pluggable is predicted, only to be pushed out to a future generation. The Rump Session will discuss if, and when, the pluggable paradigm will run out of steam, and if it’s replacement will be co-packaging.

Questions for Discussion:

  • At which switch node will pluggable no longer be a viable paradigm: 25.6T, 51.2T, 102.4T, 204.8T?
  • How important are traditional pluggable advantages of partitioning, testability, field installation and replacement, and upgradability?
  • Does an interposer between co-packaged optics and MCM preserve some of the traditional pluggable advantages?
  • What is the optimum modulation format for co-packaged optics? Is it the same or different than for pluggable?
  • Is optical link interoperability between co-packaged and pluggable optics a requirement? What if this limits co-packaged approaches?
  • It is generally assumed that co-packaging enables replacing VSR with XSR SerDes. This may result in at most 50% SerDes power savings, which translates to 20-30% host card power savings. Does this justify an entirely new paradigm? Or is at least 50% power savings required?
  • Does co-packaging enable new network paradigms, like higher radix?
  • Do advanced copper I/O techniques like fly-over cables and high-density connectors extend the pluggable paradigm?
  • An intermediate step is on-board optics. A shared limitation is that the PCB routing of on-board modules requires VSR SerDes and therefore does not reduce power. Are there on-board optics approaches that would remove this limitation and postpone the need for co-packaging?
  • Are there security advantages associated with co-packaging?
  • Which is the preferred fiber attach; vertical or edge coupled? Should it be pig-tailed or connectorized at the MCM?
  • Is MCF (multi-core fiber) required to overcome the spatial density limitations of routing out fibers from an MCM?

Format:

  • Short introductory presentations by session organizers.
  • One content slide plus one punch line slide each from a panel of industry provocateurs, adding up to 50% of session time.
  • Vigorous audience participation after each provocateur presentation, with organizers facilitating wide ranging discussion, adding up to the other 50% of session time.
  • Attendees come prepared with tough questions, insightful comments, and different perspectives to challenge the provocateurs and broaden the discussion.

Low-Power Data Center Transponders Enabled by Micrometer-scale Plasmonic Modulators (W1D.1) - Invited
08:00 AM - 08:30 AM
Plasmonic modulators allow for high-speed data modulation beyond 200GBd at the micrometer-scale and low driving voltages below 700mV. The compact footprint enables dense integration and makes plasmonic modulators a promising solution for next-generation optical interconnects.
Authors: B. Baeuerle, W. Heni, C. Hoessbacher, Y. Fedoryshyn, A. Josten, U. Koch, C. Haffner, T. Watanabe, J. Leuthold, ETH Zurich, Zurich, SWITZERLAND|B. Baeuerle, W. Heni, C. Hoessbacher, Polariton Technologies Ltd., Zurich, SWITZERLAND|C. Uhl, M. Möller, Chair of Electronics and Circuits, Saarland University, Saarbrücken, GERMANY|H. Hettrich, M. Möller, MICRAM Microelectronic GmbH, Bochum, GERMANY|D.L. Elder, L.R. Dalton, Department of Chemistry, University of Washington, Seattle, Washington, UNITED STATES|C. Haffner, Physical Measurement Laboratory, National Institute of Standards and Technology, Gaithersburg, Maryland, UNITED STATES

Distortion-aware 2D soft decision for VCSEL-MMF optical PAM interconnection (W1D.2)
08:30 AM - 08:45 AM
A distortion-aware 2D soft decision method of PAM signals have been proposed for VCSEL-MMF interconnection system. Improvements and application potential have been experimentally investigated on a 112-Gbps optical PAM-4/8 system using a multimode VCSEL.
Authors: L. Sun, J. Du, W. Zhang, Z. He, Shanghai Jiao Tong University, Shanghai, CHINA|L. Sun, C. Lu, Hong Kong Polytechnic University, Kowloon, HONG KONG|N. Chi, Fudan University, Shanghai, CHINA

168Gbps PAM-4 Multimode Fiber Transmission through 50m using 28GHz 850nm Multimode VCSELs (W1D.3)
08:45 AM - 09:00 AM
We experimentally demonstrate PAM-4 data rates beyond 160Gbps over 50m OM5 using unpackaged 850nm VCSELs. Power penalties of PAM-4 are examined demonstrating maximum data rates, with and without FEC, over 50m and 100m of fiber.
Authors: J. Lavrencik, S. Varughese, S.E. Ralph, Georgia Institute of Technology, Atlanta, Georgia, UNITED STATES|N. Ledentsov Jr., L. Chorchos, N. Ledentsov, VI Systems GmbH, Berlin, GERMANY|N. Ledentsov Jr., L. Chorchos, Warsaw University of Technology, Warsaw, POLAND

4×56-GBaud PAM-4 SDM Transmission Over 5.9-km 125-μm-Cladding MCF Using III-V-on-Si DMLs (W1D.4)
09:00 AM - 09:15 AM
We demonstrate 4×56-GBaud PAM-4 signals over 125-μm-cladding, 4-core fiber by simultaneous, direct modulation of four 1.3-μm membrane III-V-on-silicon lasers, each requiring <25-mWatts (@12 mA). A reach extension of ~15x is achieved compared to previous works.
Authors: N. Diamantopoulos, H. Nishi, T. Fujii, K. Shikama, K. Takeda, T. Kakitsuka, S. Matsuo, NTT Device Technology Labs, NTT Corporation, Atsugi, Kanagawa, JAPAN|T. Matsui, K. Nakajima, NTT Access Networks Service Systems Labs, NTT Corporation, Tsukuba, Ibaraki, JAPAN|T. Kakitsuka, Graduate School of Information, Production and Systems, Waseda University, Kitakyushu, Fukuoka, JAPAN

1.12 Tbit/s fiber vector eigenmode multiplexing transmission over 5-km FMF with Kramers-Kronig receiver (W1D.5)
09:15 AM - 09:30 AM
We demonstrate a 1.12 Tb/s MIMO-free vector eigenmode multiplexed signal transmission over 5-km 4-mode few-mode-fiber using HE11 and EH11 vector modes, 5 wavelengths and 28 GBaud 16-QAM signal with direct-detection Kramers-Kronig receiver.
Authors: J. Zhang, X. Wu, L. LU, C. Lu, The Hong Kong Polytechnic University, Hong Kong, HONG KONG|J. Li, Guangdong University of Technology, Guangzhou, CHINA|J. Tu, Jinan University, Guangzhou, CHINA|Z. Li, Sun Yat-sen University, Guangzhou, CHINA

Single λ 500-Gbit/s PAM Signal Transmission for Data Center Interconnect Utilizing Mode Division Multiplexing (W1D.6)
09:30 AM - 09:45 AM
Single wavelength 502.5-Gbit/s MDM-PAM-6 signal transmission over 20-m OM2 fiber with BER below HD-FEC threshold (3.8×10-3) is demonstrated for 400-G Data Center Interconnect without DSP for mode de-multiplexing. This scheme shows good potential for future 800-G/1.6-T DCI.
Authors: F. Li, D. Zou, Sun Yat-Sen University, Guangzhou, Guangdong, CHINA


FOSphere: A Scalable and Modular Low Radix Fast Optical Switch Based Data Center Network (W1F.1)
08:00 AM - 08:15 AM
We propose a novel scalable and modular low-radix fast optical switch based DCN with sphere topology (FOSphere). Numerical analyses on 10880-server indicates that FOSphere achieves 4.1 μs server-to-server latency and 2.6E-3 packet loss at load 0.4.
Authors: F. Yan, E. Kahan, X. Guo, F. Wang, B. Pan, X. Xue, S. Zhang, N. Calabretta, Technology University of Eindhoven, Eindhoven, NETHERLANDS

High-Throughput Optical Circuit Switch for Intra-Datacenter Networks Based on Spatial Super-Channels (W1F.2)
08:15 AM - 08:30 AM
We propose a novel optical circuit switch architecture based on spatial super-channels. We construct part of a 1,536×1,536 optical switch and its performance is experimentally confirmed. The total throughput of the switch reaches 2.1 Pbps.
Authors: E. Honda, Y. Mori, H. Hasegawa, Nagoya University, Nagoya, JAPAN|K. Sato, The National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, JAPAN

Scaling PULSE Data Center Network Architecture and Scheduling Optical Circuits in Sub-microseconds (W1F.3) - Invited
08:30 AM - 09:00 AM
PULSE, an optical circuit switched data center network, employs custom ASIC schedulers to reconfigure circuits in 240 ns. The revised PULSE architecture scales to 10,000s blades, achieves >95% sustained throughput, with low median 1.23μs and tail 145μs latencies, while consuming 115pJ/bit and costing $9.04/Gbps.
Authors: J.L. Benjamin, G.S. Zervas, University College London, London, UNITED KINGDOM

A 25.6 Tbps capacity 1024-port Hipoλaos Optical Packet Switch Architecture for disaggregated datacenters (W1F.4)
09:00 AM - 09:15 AM
We demonstrate experimentally the feasibility of a 25.6Tb/s capacity Hipoλaos optical packet switch architecture with 1024 in/out ports operating at 25Gb/s, presenting successful contention resolution and error-free operation with a control plane latency of 97.28ns.
Authors: N. Terzenidis, A. Tsakyridis, G. Giamougiannis, M. Moralis-Pegios, Informatics, Aristotle University of Thessaloniki, Thessaloniki, GREECE|N. Terzenidis, A. Tsakyridis, G. Giamougiannis, M. Moralis-Pegios, K. Vyrsokinos, N. Pleros, Center for Interdisciplinary Research & Innovation, Thessaloniki, GREECE|K. Vyrsokinos, N. Pleros, Physics, Aristotle University of Thessaloniki, Thessaloniki, GREECE

Experimental Assessments of a Flexible Optical Data Center Network Based on Integrated Wavelength Selective Switch (W1F.5)
09:15 AM - 09:30 AM
A novel bandwidth-reconfigurable optical DCN exploiting photonic-integrated WSS is experimentally assessed. Results show that optical bandwidth can be automatically reallocated according to the traffic patterns with 1.75µs end-to-end latency and 0.015 packet-loss at 0.6 load.
Authors: X. Xue, K. Prifti, B. Pan, F. Yan, F. Wang, X. Guo, N. Calabretta, Eindhoven University of Technology, Eindhoven, NETHERLANDS|F. Nakamura, H. Tsuda, Keio University, Yokohama, JAPAN

Beyond Edge Cloud: Distributed Edge Computing (W1F.6) - Invited
09:30 AM - 10:00 AM
High bandwidth demands combined with low latency applications lead the move from centralized cloud to distributed Edge Computing. We discuss how this paradigm shift impacts network interconnects design and the key network features to truly enable 5G and beyond.
Authors: N.D. Benzaoui, Nokia Bell Labs France, Nozay, FRANCE


More Than Communications: Environment Monitoring Using Existing Optical Fiber Network Infrastructure (W3G.1) - Invited
02:00 PM - 02:30 PM
We propose reusing existing optical cables in metropolitan networks for distributed sensing using a bidirectional, dual-band architecture where communications and sensing signals can coexist with weak interaction on the same optical fiber.
Authors: Y. Aono, NEC Corporation, Abiko, CHIBA, JAPAN|E. Ip, P. Ji, Optical Networking and Sensing, NEC Laboratories America, Princeton, New Jersey, UNITED STATES

Automated Thermal Drift Compensation in WDM-based Silicon Photonic Multi-Socket Interconnect Systems (W3G.2)
02:30 PM - 02:45 PM
We present an on-chip AWGR-based interconnect system with automated thermal drift compensation along cascaded resonant structures in a dual-socket layout. Error-free operation in a 30 Gb/s data-routing scenario within a 12C temperature range is demonstrated.
Authors: M. Moralis-Pegios, K. Fotiadis, S. Pitris, T. Alexoudi, N. Pleros, Aristoteleio Panepistimio Thessalonikis, Thessaloniki, GREECE|F. Zanetto, E. Guglielmi, V. Grimaldi, D. Aguiar, G. Ferrari, M. Sampietro, A. Melloni, Dipartimento di Elettronica Informazione e Bioingegneria, Politecnico di Milano, Milan, ITALY|P. De Heyn, Y. Ban, J. Van Campenhout, imec, Leuven, BELGIUM

BER and TDECQ Correlation for Different Impairments in 400Gbps PAM4 system (W3G.3)
02:45 PM - 03:00 PM
Closed-form bit-error rate (BER) expression as a function of transmitter dispersion eye closure quaternary (TDECQ) is derived. Based on a silicon-photonics 400-Gbps PAM4 transceiver, BER and TDECQ correlation is verified for different impairments.
Authors: Y. Zhao, C. Doerr, L. Chen, N. Zhu, D. Ton, R. Aroca, X. Huang, M. Xu, Acacia Communication Inc., Holmdel, New Jersey, UNITED STATES

A 0.57-mW/Gbps, 2ch x 53-Gbps Low-Power PAM4 Transmitter Front-End Flip-Chip-Bonded 1.3-μm LD-Array-on-Si (W3G.4)
03:00 PM - 03:15 PM
A low-power 2-channel PAM4 transmitter front-end consisting of 65-nm CMOS PAM4 shunt LD drivers and flip-chip-bonded 1.3-μm LD-array-on-Si achieves simultaneous 2ch x 53-Gps PAM4 transmission over 2-km-long SSMF with power efficiency of 0.57 mW/Gbps.
Authors: T. Kishi, M. Nagatani, K. Shikama, T. Fujii, H. Nishi, H. Yamazaki, N. Sato, H. Nosaka, S. Matsuo, NTT Device Technology Laboratories, Atsugi-shi, JAPAN|S. Kanazawa, NTT Device Innovation Center, Atsugi-shi, JAPAN

The Role of Optics In Future AI-driven Intra-DC Infrastructure (W3G.5) - Invited
03:15 PM - 03:45 PM
The next generation of artificial intelligence and machine learning requires the ability to connect multiple nodes across an ever-increasing scale. This growth is driving an increased role of optics to build these next generation system.
Authors: B. Booth, Microsoft Corp, Redmond, Washington, UNITED STATES


Session Description

As optical connectivity migrates from 100Gbps NRZ to 200Gbps and 400Gbps PAM-4 based solutions, optical module users are faced with increased power consumption, latency and cost challenges. To addresses these challenges, multi-disciplinary industry leaders formed OpenEye MSA which expands optical module technology solutions to include analog CDR and analog PLL to provide non-DSP-based alternative. This panel will discuss the trade-offs and interoperability between the two approaches, and benefits of this new solution for low cost and low power high speed interconnect for data centers and campus networks. Subjects covered will include available analog semiconductor technologies, test solutions, emerging OpenEye compliant single mode and multimode optical modules and OpenEye MSA roadmap.

Moderator

Marek TlalkaSenior Director of Marketing, High Performance Analog, MACOM, USA 

Panelists

  • Benny Koren, VP of Architecture, Mellanox, USA
  • Greg LeCheminantMeasurement Applications Specialist, Digital Communications Analysis, Keysight Technologies, USA 
  • Jeffery Maki, Distinguished Engineer II, Juniper Networks, USA
  • Marek TlalkaSenior Director of Marketing, High Performance Analog, MACOM, USA 
  • Tim Vang, Vice President, Marketing and Applications, Semtech Corporation, USA 
  • Michael XinVice President, Sales and Marketing, Cambridge Industries USA Inc. (CIG), USA

Moderator: Loukas Paraschis, Senior Director, Infinera, USA

Panel Description:

The networks interconnecting datacenters, often referred to as Data Center Interconnects (DCIs), have grown more than any other transport infrastructure in the last decade, and DCI’s are projected to grow by several orders of magnitude in the coming years.  The economics associated with DCI growth have motivated: 1) dedicated DCIs accounting for some of the most spectrally efficient fiber networks deployed globally, and 2) a new class of purpose-built systems optimized for throughput and power consumption.  The emergence of DCIs has led to extensive adoption of software innovations in automation, programmability, management abstraction, and control-plane disaggregation, as well as the introduction of the associated open transport architectures.   

This panel will discuss the extent to which these innovations have been adopted in current and future DCI networks.  For example, 400GE coherent WDM pluggables in DCI routers have been advocated, often combined with system and control-plane disaggregation, in open (multi-vendor) network architectures that minimize operational overhead without compromising significantly transport network performance. At the same time, the potential value from optimization and traffic engineering, such as from network analytics, potentially even combined with optical margin “harvesting” or restoration in addition to IP/MPLS protection, has also recently been proposed.   

This panel will introduce the views of several industry leaders in this area and evoke discussion among panelists and the floor. 

Speakers:

  • Andreas BechtolsheimFounder, Chairman and Chief Development Officer, Arista Networks, USA
  • Nancy El-SakkaryStaff Network Engineer, Google, USA
  • Michael Strunz-KrollHead of Architecture and Engineering, euNetworks, Germany
  • Chongjin XieSenior Director, Alibaba Group, USA
  • Yawei YinPrincipal Network Developer Manager, Microsoft, USA
    Getting More into Less: A Cloud Provider’s View of the Future of DCI Optical Network.

Moderator: Hideki Isono, Senior Professional Engineer, Fujitsu Optical Components, Japan

Panel Description:

Global Datacenter IP traffic, growing rapidly at a 25 percent year over year growth rate, is to become 20 Zettabytes in 2021, according to Cisco Global Cloud Index 2018. To meet this huge capacity demand, high speed network systems are being deployed with optics operating at 100Gbit/s and beyond.  

The IEEE802.3 has completed new Ethernet standards for rates up to 400Gbit/s (802.3bs/cd) in 2018 adopting 25GBaud PAM4 50Gbit/s optical modulation (8x50G). They have also recently introduced the 50GBaud PAM4 100Gbit/s optical modulation format as new Ethernet technology building blocks for 8x50G and 4x100G 400GE optical interfaces. As for electrical interfaces, 50Gbit/s per lane 400GE has already matured (8x50G) and 100Gbit/s per lane for 4x100G is being discussed. 

Along with these developments, new pluggable form factor MSAs such as QSFP-DD and OSFP have been commercialized and matured. As a result, maximum system board-level data throughputs have reached as high as 14.4Tbit/s. 

The next phase of the datacenter technology roadmap (beyond 400G) remains somewhat unclear. Solutions for 800G and 1.6T can be expected; however, it will take several years for technologies for 100Gbit/s electrical and optical lane rates to achieve desirable power consumption and density targets.  If industry sticks to pluggable solutions, we will see parallel solutions such as dual- or quad- 400G pluggable modules. This will require drastic improvements in integration technologies (see the optical integration focused panel at this OFC) and improvements in CMOS processes below 7nm to restrain increases in power consumption density. An alternative direction would be on-board optics and co-packaging, which may need new technology breakthroughs and more time to mature. 

This panel will introduce the views of several industry leaders in this area and evoke discussion among panelists and the floor. 

Speakers:

  • Andreas BechtolsheimChairman and Chief Development Officer, Arista Networks, USA
    The Road to 800G

    As the demand for more bandwidth in the cloud keeps growing, the most cost-effective solution for data center switching and optics vendors is to embrace the 100G SERDES ecosystem and 800G optics. In my presentation, I will discuss the status of 800G data center optics and system developments together with some innovations that will facilitate the adoption of these 800G optics in existing data centers.
  • Robert BlumDirector of Marketing and New Business, Intel, USA
    Integrated Silicon Photonics for High-volume Data Center Applications

    The rapid growth in data center traffic has been the key driver for the increased performance of networking equipment, such as Ethernet switches and optical pluggable transceivers. But bandwidth scalability challenges around density, cost, and power will require much tighter integration of optics and networking silicon in the near future. We will discuss how silicon photonics combined with advanced packaging approaches can enable higher density, reduced power per bit, and ultimately the continued scalability of network bandwidth and performance.
  • Brad Booth, Principal Hardware Manager, Next Cloud System Architecture, Microsoft, USA
    Network Diversity

    Machine learning, artificial intelligence, front end network, backend network, edge, core, DCI, etc. Equipment and transceiver manufacturers look to find commonality to develop products for market volume, but the list is long on what drives differentiation. Is there commonality to be found or will diversity persist?
  • Jeffery MakiDistinguished Engineer II, Juniper Networks, USA
    Bandwidth Density Progression

    Bandwidth evolution of routing and switching systems have converged at 12.8 to 14.4 Tb/s capacity per 1RU with adoption of 400 Gb/s Ethernet in a common form factor. There are competing points of view about how to facilitate higher bandwidth density. The presenter will review factors that influence the evolution and choice of implementation.
  • Mark Nowell, Fellow, Cisco, USA
    400 GbE Hits the Mainstream Inside the Data Center

    With years of development and anticipation leading up to the availability of 400 GbE, deployments are now in progress.  This talk will focus on how the industry focus on key technology building blocks will lead 400 GbE to be broadly deployed and then lead to what comes next.
  • Sang-Yoon (Sy) RheeMarketing Director, COO, Fujitsu Optical Components America, Inc. (FOCUS), USA
    Technology Options for Next-Gen Data Center Optics

    This presentation discusses the challenges and solutions towards the development of optical modules that are required for high speed connectivity inside the data center.  Recent progress in module configuration, transmission schemes, DSP, thermal management are reviewed and their future prospects are discussed.

Session Description

400ZR is bringing the benefits of coherent pluggable solutions to the data center edge network environment after a relatively rapid convergence thanks to hyperscale operators, module suppliers, equipment vendors, and the Optical Internetworking Forum. The question now is, where do we go from here? This session wil discuss with industry experts the potential areas of expansion beyond 400ZR such as:  data rates beyond 400G, channel spacing evolution, technology enablers, and how these will collectively affect next generation architectures beyond DCI edge.

Moderator

Tom Williams, Vice President of Marketing, Acacia Communications, USA

Speakers

  • Martin Birk, Director, Optical Platforms, AT&T, USA
  • Jörg-Peter Elbers, SVP Advanced Technology, Standards & IPR, ADVA Optical Networking, Germany
  • Tad Hofmeister, Optical Network Architect, Google, USA

Performance Oriented DSP Design for Flexible Coherent Transmission (Th3E.1) - Tutorial
02:00 PM - 03:00 PM
We review the impact of DSP in terms of performance and flexibility in the data network. DSP has addressed the optimization of capacity against reach and power. Future DSP targets cost-reduction through flexible point-to-multi-point architectures.
Authors: C.R. Fludger, Infinera GmbH, Nuremberg, GERMANY

1.1 Tb/s/l at 9.8 bit/s/Hz DWDM transmission over DCI distances supported by CMOS DACs (Th3E.2)
03:00 PM - 03:15 PM
We report on a 16-nm CMOS DAC based transmitter optimization enabling bitrates up to 1.15 Tb/s. We successfully demonstrate DWDM transmission over DCI distances up to 118 km at 1.1 Tb/s and spectral efficiencies of 9.8 bit/s/Hz.
Authors: F. Buchali, M. Chagnon, K. Schuh, V. Aref, Nokia Bell Labs, Stuttgart, GERMANY|V. Lauinger, KIT, Karlsruhe, GERMANY

Maximizing Throughput via Vertical Optimization of the Coherent MODEM (Th3E.3) - Invited
03:15 PM - 03:45 PM
Vertical optimization of DSP algorithms, analog electronics, optical components and PCB design is critical to maximize the SNR limit of the digital coherent MODEM. We demonstrate a record net ISD of 10.82b/s/Hz for a vertically optimized 256QAM transceiver operating at a symbol rate >50GBd
Authors: R. Maher, A. Nguyen, Z. Wang, S. Koenig, M. Missey, A. Le Liepvre, R. Going, S. Wolf, P. Samra, P. Day, M. Ziari, F. Kish, S. Sanders, P. Kandappan, Infinera Corporation, Sunnyvale, California, UNITED STATES|M. Torbatian, S. Tremblay, Infinera, Ottawa, Ontario, CANADA


Performance Oriented DSP Design for Flexible Coherent Transmission (Th3E.1) - Tutorial
02:00 PM - 03:00 PM
We review the impact of DSP in terms of performance and flexibility in the data network. DSP has addressed the optimization of capacity against reach and power. Future DSP targets cost-reduction through flexible point-to-multi-point architectures.
Authors: C.R. Fludger, Infinera GmbH, Nuremberg, GERMANY

1.1 Tb/s/l at 9.8 bit/s/Hz DWDM transmission over DCI distances supported by CMOS DACs (Th3E.2)
03:00 PM - 03:15 PM
We report on a 16-nm CMOS DAC based transmitter optimization enabling bitrates up to 1.15 Tb/s. We successfully demonstrate DWDM transmission over DCI distances up to 118 km at 1.1 Tb/s and spectral efficiencies of 9.8 bit/s/Hz.
Authors: F. Buchali, M. Chagnon, K. Schuh, V. Aref, Nokia Bell Labs, Stuttgart, GERMANY|V. Lauinger, KIT, Karlsruhe, GERMANY

Maximizing Throughput via Vertical Optimization of the Coherent MODEM (Th3E.3) - Invited
03:15 PM - 03:45 PM
Vertical optimization of DSP algorithms, analog electronics, optical components and PCB design is critical to maximize the SNR limit of the digital coherent MODEM. We demonstrate a record net ISD of 10.82b/s/Hz for a vertically optimized 256QAM transceiver operating at a symbol rate >50GBd
Authors: R. Maher, A. Nguyen, Z. Wang, S. Koenig, M. Missey, A. Le Liepvre, R. Going, S. Wolf, P. Samra, P. Day, M. Ziari, F. Kish, S. Sanders, P. Kandappan, Infinera Corporation, Sunnyvale, California, UNITED STATES|M. Torbatian, S. Tremblay, Infinera, Ottawa, Ontario, CANADA


Sponsored by: