Standards, Technology, Market and Industry Trends for MMFG. Mabud Choudhury, Robert Lingle, Jr., John Kamino, Roman Shubochkin,

Durgesh Vaidya, David Braganza, Yi Sun OFS, 2000 NE Expressway, Norcross, GA 30071

ysun@ofsoptics.com

Abstract We present the latest standards, technology, market and industry trends for graded-index, laser optimized multimode fibers (MMF) coupled to low-cost Vertical-Cavity Surface-Emitting Lasers (VCSELs) links, which have historically provided the most cost-effective and widely deployed short-reach fiber solutions for local area networks (LAN) and data centers (DC). An updated outlook on the future of MMF will be presented.

We conducted this study by analyzing, evaluating and summarizing key recent industry updates, market forecasts, customer needs, standardization projects and latest fiber, connectivity, module & transceiver technologies as the demands on cloud/web 2.0/hyperscale, enterprise, telecom/central office, co-location and edge DCs continue to evolve. The last few years have seen strong activity in fiber (TIA and IEC), cabling (ANSI/TIA and ISO/IEC) and application (IEEE 802.3 Ethernet and INCITS T11 Fibre Channel) standards – as OM5 fiber was defined; as Ethernet speeds move from 10GbE/40GbE/100GbE to 25GbE/50GbE/400GbE and beyond, 800GbE/1.6TbE; and as Fibre Channel speeds move from 8GFC/16GFC to 32GFC/64GFC/128GFC (serial/duplex) and 128GFC/256GFC (parallel) and beyond 512GFC (parallel). At the same time, VCSEL transceiver technologies have begun moving from Non-Return to Zero (NRZ) to Pulse Amplitude Modulation 4 (PAM4) modulation, and providing single and multiple wavelength solutions - shortwave wavelength division multiplexing (SWDM) and bidirectional (BiDi)/codirectional (CoDi) solutions to support increasing data rates. The strong adoption of 100GBASE-SR4, the continued success of 40G-BiDi duplex solutions, the introduction of 40G-SWDM4, 100G-SWDM4 and 100G-BiDi solutions along with the recently formed 400G-BiDi multi-source agreement (MSA) group underscore the ongoing success and continued relevance of MMF links. A major focal point in defining and developing the future of MMF is the recently formed IEEE P802.3cm “400 Gb/s over Multimode Fiber Task Force” to leverage and address these market and technology trends, underscoring the industry commitment and continued end-user demand for MMF/VCSEL solutions.

The paper shows the continued relevance of MMF links as cost-effective, short-reach solutions as driven by latest standards, technology and market trends. The reasons VCSEL-based transceivers can maintain a cost advantage will be discussed, even as speeds increase and single-mode fiber transceivers move into the 500m space in data centers in higher volumes. Our key conclusions are that cloud and enterprise data centers continue to be a strong demand base for short reach, cost effective multimode solutions; in addition, off-premises colocation, close to end customer fog and edge DCs and telecom central office/DCs provide new and growing addressable markets for the cost/power advantages of short reach <100 m / 150m OM3/OM4/OM5-VCSEL links. These market and customer needs are supported by new application standards from IEEE 802.3 and T11 Fibre Channel, along with MSA and multi-vendor proprietary solutions.

Keywords: Multimode Fiber; MMF; Laser-Optimized Multi-mode Fiber; LOMMF; Vertical-cavity surface-emitting laser; VCSEL; Ethernet; Fibre Channel; TIA; ISO; IEC; Data Center; DC; Enterprise; Cloud; OM3; OM4; OM5; Shortwave Wavelength Division Multiplexing; SWDM; Bi-Directional; BiDi; Pulse Amplitude Modulation; PAM4; 100Gb/s, 200 Gb/s; 400 Gb/s.

1. Introduction 1.1 Key Advantages of MMF & MM VCSELs The key enduring advantages of VCSELs over Light Emitting Diode (LED) and Edge Emitting Laser (EEL) technology are as follows: manufacturability, integration, reliability, testability, packaging, custom packaging, low power, and power efficiency [1]. MMF with a larger core size than Single-mode Fiber (SMF) can guide multiple modes and can more easily capture light from a transceiver thereby decreasing alignment costs; the less stringent alignment costs for MMF relative to SMF also extend to having lower cost multimode connectors. While MMF fiber cost is higher than SMF, for <100 m and even <500 m short reaches, the lower cost connectivity, installation and ongoing maintenance/usability along with lower cost (than the still relatively higher cost SM optical sources, which have declined due to Silicon Photonics and other technologies) optical source and optoelectronics of 850 nm MM VCSELs makes the Laser Optimized MMF (LOMMF)/VCSEL combination the most cost effective short-reach solution for traditional on-premises enterprise applications, emerging off-premises applications, as well as several generations of hyperscale cloud data centers.

2. Current Market & Future Trends 2.1 Ethernet Market The Ethernet market continues to grow and move to higher speeds. For traditional on-premises Enterprise Data Center (DC) and Local Area Network (LAN) applications, OM3/OM4 and MM VCSEL-based optics dominate structured cabling and point-to-point short-reach links. The vast majority of traditional Enterprise links are <100 m, while <300 m covers virtually all traditional Enterprise optical data center links. For speeds ≤10Gb/s, including 1GbE and 10GbE, 1000BASE-T and 10GBASE-T, along with newer 2.5GBASE-T and 5GBASE-T twisted pair solutions have had remarkable success. At speeds >10Gb/s, Active Optical Cables (AOC) are typically used for <30 m links and Direct-Attach Copper (DAC) are often used for <5 m reaches. 10GBASE-SR SFP+ modules serial solutions and 40GBASE-SR4 QSFP/QSFP+ module parallel/4-fiber pair solutions have been widely deployed. Standards+ and proprietary solutions such as 40GBASE-eSR4 provide extended reach (reaches up to 400 m over OM4 fiber), and 40G BiDi, first multi-wavelength MM solution is widely deployed to support single pair transmission. The 25GBASE-SR standard was completed in 2016. 100GBASE-SR4 is being deployed in hyperscale and leading edge enterprise data centers.

The IEEE 802.3 Working Group that develops standards for Ethernet networks completed IEEE 802.3bs in 2017, which defined 400GBASE-SR16 VCSEL-MMF links over 25 GBaud lanes on OM3/OM4/OM5 (operating at 850 nm), and is actively developing 50GBASE-SR, 100GBASE-SR2 and 200GBASE-SR4 on 50 GBaud lanes over OM3/OM4/OM5 (operating at 850 nm) standards in P802.3cd Task Force. In 2017, the Shortwave Wavelength Division Multiplexing (SWDM) Multisource Agreement (MSA) released specifications for 40G SWDM4 and 100G SWDM4 with commercial products released by the end of 2017; 100G BiDi solutions were introduced in the same timeframe – expanding the options for 40GbE and adding 100GbE MMF/VSCSEL solutions for duplex fiber, single-pair links.

In 2018, IEEE 802.3 WG formed the IEEE P802.3cm 400 Gb/s over Multimode Fiber Task Force, which is developing standards for 400GBASE-SR4.2 and 400GBASE-SR8, both based on 50 Gb/s lanes on OM3/OM4/OM5, with SR4.2 operating at 850 nm & 910 nm, and SR8 operating at 850 nm. This will be the first standards based application to take advantage of OM5 fiber’s wideband capabilities to support multi-wavelength solutions. A 400G BiDi MSA was also formed in 2018 – another multi-wavelength VCSEL-MMF application. IEEE 802.3 WG also authorized the IEEE P802.3ck 100 Gb/s, 200 Gb/s, and 400 Gb/s Electrical Interfaces Task Force to generate 100 Gb/s per electrical lane standard.

Figure 1 shows Ethernet Alliance’s The 2018 Ethernet Roadmap [2] for Ethernet speeds documenting the growth in Ethernet speeds to satisfy the market needs of hyperscale cloud and largest Enterprise data center customers.

Figure 1. Ethernet Alliance’s The 2018 Ethernet

Roadmap for Ethernet Speeds [2]

2.2 Fibre Channel Market On the storage side, Fibre Channel (FC) remains the dominant standard for on-premises enterprise networks. Over 90% of Fibre Channel links utilize MMF and VCSELs [3]. Dell’Oro [4] estimates that 113 million ports of FC switches and adapters have shipped since 2001 and estimate that 46.2 million ports are still currently in operation with continued growth of FC links [5]; 16GFC is being deployed at faster rate than 8GFC and 32GFC is beginning to ramp up [5]. INCITS T11 Technical Committee is currently developing serial lane 64GFC in SFP and 256GFC parallel lane in QSFP, and has recently approved the project to develop 128GFC serial lane. Fibre Channel has also expanded to Fibre Channel over Ethernet (FCoE) allowing FC protocols to run over Ethernet networks, and is

completing development of FC standard that maps FC to NVMe over Fabrics, allowing increased network performance and lower latency [5]. Figure 2 shows Fibre Channel Industry Association’s (FCIA) “Roadmap for Fibre Channel”, which will continue to remain a major application for MMF and VCSELs.

Figure 2: From FCIA’s Fibre Channel Roadmap [3]

2.3 Data Center Trends The single biggest trend for data centers in the last few years has been the migration to cloud data centers at an accelerated rate. The widely cited and well regarded Cisco Global Cloud Index (GCI), 2016-2021 [6] predicts the following key items:

By 2021, 94 percent of workloads and compute instances will be processed by cloud data centers; 6 percent will be processed by traditional data centers (see Figure 3) [6]. Worldwide cloud traffic will more than triple by 2021 (Figure 3). Hyperscale data centers will grow from 338 in number at the end of 2016 to 628 by 2021. They will represent 53 percent of all installed data center servers by 2021 (see Figure 4) [6].

Figure 3: Workloads/compute instance:

Traditional DC vs. Cloud DC [6]

Figure 4: Global hyperscale data center growth [6]

2.3.1 What does “cloud” mean? The term “cloud” can cover many definitions and meanings – both technical and marketing. National Institute of Standards and

Technology (NIST) developed “The NIST Definition of Cloud Computing,” published in 2011 [7]. Cisco GCI and Figure 3 utilizes the NIST definition of cloud computing in defining cloud data centers. Per the NIST cloud computing definition: “cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g. networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.” [7] The five essential characteristics are: on-demand self-service, broad network access, resource pooling, rapid elasticity, and measured service. The three service models are: Software as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS). The four deployment models are: private cloud, community cloud, public cloud, and hybrid cloud [7]. A graphical depiction of hybrid cloud model, which is increasingly relevant for the enterprise market, is shown in Figure 5 [8].

Figure 5: Hybrid Cloud [8]

Cloud refers to the way IT resources are configured and accessed, not necessarily the size of the facility. A very wide range of data centers are moving to cloud computing. The accelerating growth of the cloud can also be viewed through the prism of private cloud dc vs. public cloud dc vs. non-cloud dc, as shown in Figure 6 [6].

Figure 6: Private vs. Public vs Non-Cloud DC [6]

The focus tends to be on public (open use by the general public) cloud data centers since these data centers have the highest growth rate and tend to be operated by the largest technology companies. However, private (for exclusive use by a single organization) cloud data centers are also projected double digit CAGR. Private cloud data centers can be on-premises or off-premises (e.g. multi-tenant data centers or co-location data centers); can include large enterprise data centers or be provisioned as a service or app by public cloud data centers. The shrinking part of the data center market is non-cloud data center.

2.3.2 What does “hyperscale” mean? Cisco GCI defines “hyperscale” criteria based on revenue: “>US$1 billion in annual revenue from Infrastructure as a Service (IaaS), Platform as a Service (PaaS), or infrastructure hosting services; >US$2 billion in annual revenue from Software as a Service (SaaS); >US$8 billion in annual revenue from e-commerce/payment processing” [9]. Twenty-four providers meet these Cisco CGI hyperscale criteria, and are the basis for Figure 4. These 24 hyperscale data center companies, which have strong and growing market presence and excellent technical capabilities, utilize a wide variety of architectures and different optical link solutions/technologies – both SMF and MMF.

Some of the largest US and China based hyperscale cloud companies are currently deploying 100G-SR4 and have 400G-SR4.2 or 400G-SR8 in their roadmap [10] [11][12][13][14]. While some of the largest hyperscale cloud providers have adopted parallel SMF Parallel Single Mode 4 lane (PSM4) solutions and some have committed to a duplex SMF strategy, others “employ a mix of copper, short reach (SR) multi-mode optics, and long reach (LR) or Coarse Wavelength Division Multiplexing (CWDM4/CLR4) single-mode optics. The most critical metrics for intra-datacenter interconnects are cost, size, and power, in order of importance. The application of the interconnect has followed this priority, where we adopt the lowest cost solution possible. Thus, copper is used for all intra-rack interconnect, SR for reaches up to 100m, and LR/CWDM4/CLR4 up to 2km..” [15] Alternate valid definitions for “hyperscale” primarily refer to the size of the physical facility, requiring longer cable runs, for example > 300 m and up to 2 km reaches; the term “mega data center” is also often used to categorize the physically largest/longest cable run data centers.

2.3.3 Enterprise Data Centers The rapid growth of cloud data centers and hyperscale data centers, along with the decline of non-cloud data centers at times leads to the misinterpretation that enterprise data centers are disappearing. The reality is that significant percentage of enterprise data centers, especially the larger ones, are being transformed to predominantly cloud data centers. They are utilizing greater server virtualization, containerization, Software Defined Networks (SDN), Network Functions Virtualization (NFV), analytics and automation to provide the essential characteristics of cloud computing: on-demand self-service, broad network access, resource pooling, rapid elasticity, and measured service.

Figure 7 shows Dell’Oro’s 5-year forecast, 2018-2022 for Ethernet switch revenue by data center segments [16]. While the switch revenue from overall enterprise market declines slightly in the 5 year forecast, the switch revenue forecast for large enterprises shows significant growth. These large enterprise data centers are adopting hybrid cloud strategies with private cloud implementations, both on-premises and off-premises, to develop the best cost effective data center solutions based on their business needs.

Source:Cisco

Figure 7: Dell’Oro Ethernet Switch Data Center Forecast [16]

Given the size/reach, architecture, data, networking and storage needs of the enterprise segment, MMF and VCSELs will remain a cost effective, low power Ethernet and Fibre Channel solution for this significant market segment.

2.3.4 Telecom/Central Office Data Centers Central Offices are being transformed by multiple industry trends: Increased bandwidth demand – next generation PON and 5G; Convergence, merging of wireline and wireless – converged access infrastructure, networking platforms and applications; SDN and NFV - key technologies that enable virtualization in data center and central office environments; Edge compute – moving computing closer to the edge for lower latency; and Network simplification [17].

A major intiative is Central Office Re-architected as a Datacenter (CORD), see Figure 8 [18]. CORD hardware architecture is based on “commodity servers interconnected by a fabric of white-box switches; switching fabric in a spine-leaf topology for optimized east-west traffic; specialized access hardware for connecting subscribers (residential, mobile and/or enterprise).” [18]

Figure 8: CORD Virtualized Central Office [18] Globally, telecommunications service providers have different initiatives for their central office transformation to data center efforts. These efforts are still early, but the market potential and market size of addressing customer needs with data center and cloud functionality are significant. The sizes of the “data centers” being built inside central offices are ideal to take advantage of the cost and power advantages of MMF/VCSEL solutions.

2.3.5 Edge Data Centers As Internet of Things (IoT) adds an exponentially growing number of devices away from centralized cloud data centers, with growing utilization of local machine learning, analytics/AI, mobile data, and applications requiring low latency, edge computing and edge data center market are growing rapidly (see Figure 9) [19].

Figure 9: Global Edge Data Center Market, 2018-2022 [19] Hyperscale cloud computing drives a centralized computing model, edge computing swings pendulum back to distributed computing (see Figure 10) [20].

Figure 10: Edge Computing Representation [20]

Currently edge computing uses a mix of wired and wireless connectivity, copper and fiber – direct attach cables (DAC), active optical cable (AOC), MMF or SMF depending on data center size and company network architecture choice. Future needs, changing architecture, and the dramatic increase of local data traffic from IoT will likely require structured cabling and reaches that continue to make MMF/VCSELs an effective option in this rapidly growing market segment with short reach requirements.

2.3.6 Co-Location Data Centers In addition to hyperscale and on-premises enterprise data centers, co-location data centers are also a growing segment of the off-premises enterprise DC, hybrid cloud solution mix. Industry leaders use a wide variety of network connectivity in their co-location DCs including MMF.

2.4 100GbE Figure 11 [21], provided by LightCounting, a market research firm that tracks the module industry, shows that for 2015-2017 100GbE QSFP28 consumption, 100GbE SR4 has similar volumes to 100GbE PSM4 and greater than the units shipped of 100GbE LR4, but less than the volumes for 100GbE CWDM4 and CLR4 [21]. Currently, data centers utilizing 100G switching are

Source:Intel

primarily Cloud DCs and some leading-edge enterprise DCs. A significant portion of these early adopters are utilizing 100GBASE-SR4 solution, underscoring the continued cost/power/overall value proposition of MMF-VCSEL solutions with increasing data rates.

Figure 11: 100GbE QSFP28 Unit Shipments, 2015-2017 [21]

Figure 12 [22] shows LightCounting’s prediction of strong 100GbE growth from 2018 to 2023 for both 100-300 m reach (comprised of 100GbE SR4, eSR4, SWDM4 and BiDi) and 100GbE with 2km reach (comprised of CWDM4/CLR4 early, and includes FR in later years) [22]. Both MMF and SMF solutions are predicted to grow strongly.

Figure 12: 100GbE by reach – Shipments, 2010-2023 [22] Uptake of 100GbE MMF modules portends need for 400GbE MMF modules in the future.

2.5 MMF Forecast The continued relevance and growth of 100 GbE MMF modules and new 400 GbE MMF standards and solutions supports the CRU projection of growth for MMF shown in Figure 13 [23].

Figure 13: Worldwide MMF Demand, 2014-2022, CRU [24] SMF modules have become increasingly more relevant in the past few years as the cost structure for SMF modules has declined and as data rates have increased. The percentage of MMF modules – once the dominant optical module type – has declined relative to the percentage of SMF modules over the past 5 years for data rates of 25G to 400G [24]. Importantly, however, note that the relative decline of MMF modules in comparison to SMF modules does not mean that MMF module shipments are declining. Indeed, as shown in Figure 14 [24], a LightCounting forecast, there is strong projected growth for MMF modules, along with the growth of SMF modules. In fact, the share of MM modules is projected to increase from around 35% in 2016 to 40% in 2021. The future for MMF modules continues to be a positive one.

Figure 14: MMF vs SMF 25G to 400G modules, 2012-2023

[24]

3. Standards Ecosystem and Update 3.1 Standards Ecosystem Enterprise customers, vendors and the overall industry have benefited from interoperable, multivendor industry standards developed by the interaction and coordination between TIA and IEC component/fiber standards, ANSI/TIA and ISO/IEC structured cabling standards and IEEE and Fibre Channel application standards. OM3 growth coincides with 10GBASE-SR and 4GFC and 8GFC. OM4 supports extended reach for prior applications and supports 40GBASE-SR4/100GBASE-SR10/100GBASE-SR4 and 16GFC/10GFCoE/32GFC/128GFC for extended reach than OM3. As the newer applications are deployed, OM3/OM4 sales continue to grow. The development of OM5 creates a fully equivalent and backward compatible MMF to OM4 and provides a foundation for the more effective implementations of multiple wavelengths via

SWDM or BiDi technologies for higher data rates over fewer fiber pairs to support data rates of up to 400 Gb/s and beyond.

3.2 Recent TIA & IEC & and ANSI/TIA & ISO/IEC Standards for MMF Within the last few years, TIA & IEC and ANSI/TIA & ISO/IEC developed the OM5 specification that extends the 850nm performance of OM4 out to 953nm to take advantage of SWDM and BiDi VCSEL technologies. Recently, these standards bodies have initiated efforts to characterize OM3 and OM4 between 850 and 953 nm to further support multi-wavelength applications and solutions. During April 2018 meeting in Germany, IEC Subcommittee 86A on “Fibres and Cables” accepted US delegation comments to the 6th Edition of the IEC 60793-2-10 document that contains specification for all category A1 multimode fibers. One of the accepted comments defined EMB guidance for OM3/4 MMF at wavelengths between 850 and 953 nm (see Fig. 15). The EMB guidance for OM3/4 fibers will support future work of the recently formed IEEE P802.3cm 400 Gb/s over Multimode Fiber Task Force. During October 2018 meeting in Busan, S. Korea, the new 7th edition of the 60793-2-10 document successfully closed Committee Draft for Vote (CDV); the standard will be completed within first half of 2019. In the meantime, the TIA TR42 “Telecommunications Cabling Systems” Committee and its TR42.12 subcommittee on “Optical Fibers and Cables” approved project authorizations for adaption of IEC’s fiber specifications including: IEC 60793-2 (General) to become ANSI/TIA-4920000-C and IEC 60793-2-10 (MMF family) to become ANSI/TIA-492AAAF. The latter document will replace current separate documents for separate MMF types (TIA-492AAAA, AAAB, AAAC, AAAD, AAAE) and include all of them in one document. The OM3 and OM4 bandwidth information for SWDM will be added to the 492AAAF document as a regional difference while IEC works to revise their MM specs in parallel.

Figure 15: EMB guidance for OM3/4 MMF between 850

and 953 nm according to the IEC 60793-2-10 ed.7 Committee Draft for voting. TIA guidance for OM5 is

shown for comparison. 3.3 Recent Fibre Channel and FCIA for MMF INCITS T11 FC-PI-7/64GFC (serial) completed first Letter Ballot vote in July 2018 with unanimous approval and some comments. All technical work was completed in 2018, and final public review has also been completed. It is expected that FC-PI-7/64GFC standard, which specifies 100 m reach for OM4/OM5

and 70m reach for OM3, will become a published standard by June 2019.

The next open T11 project is FC-PI-7P/256GFC (parallel), which also has target reaches of 100 m and 70 m for OM5/OM4 and OM3, respectively. Current proposal, as of February 2019, is to include only MMF tables from FC-PI-7. The parameters in the tables will be unchanged from 64GFC. The distance will be 100 meters on OM4/OM5. Some additional margin analysis work has to be completed before the final decision can be made. If the MMF table can be left unchanged then 256GFC will be done as an amendment to FC-PI-7; so there would be no separate FC-PI-7P standard.

A new project proposal for FC-PI-8/128GFC (serial/single lane) was initiated in December, 2017 and subsequently became an approved project in April, 2018. The FCIA Marketing Requirements Document (MRD) for FC-PI-8 and the project proposal were reviewed in February 2018. The FC-PI-8 MRD has target technical stability date of 2021 and product availability of 2022. The requirements include backward compatibility to 32GFC/64GFC and 100 m reach for OM5/OM4. This project will leverage the work in IEEE P802.3ck 100 Gb/s, 200 Gb/s, and 400 Gb/s Electrical Interfaces Task Force for the electrical lane specifications, the work in OIF for 100 Gb/s optical lane work and, potentially, the work in IEEE P802.3cm. Key technical challenges such as technical feasibility of 100 Gb/s VCSELs within targeted project completion dates, meeting backplane loss budget and handling host PCB loss remain to be solved. During February 2019 Fort Worth, TX, US FC-PI-8 Working Group ad hoc meeting, key contributions were made on: 128GFC chip to module channels – based on first pass simulations (not final proposal) a line rate of 112.2Gbps is possible (for comparison, line rate for Ethernet is 106.25Gbps, 128GFC MRD is 115.6Gbps); 128GFC speed negotiation - use 28GBaud TTS for 32/64/128GFC, extending what was done for 64GFC to 128GFC; Clock content in PAM4 data streams for FC protocol; and FC-PI-8 128GFC MMF solutions – based on straw poll, there was unanimous support for a true serial 128GFC variant, a multi-wavelength variant, 56Gb/s BiDi solution and a multi-wavelength variant, 56Gb/s CoDi solution. Meeting the target completion dates and all requirements per FC-PI-8 MRD will be very challenging.

3.4 Recent IEEE for MMF 3.4.1 802.3bs and 802.3cd Wideband MMF and TIA-492AAAE (operating at 850 nm) references were added in September, 2016 to IEEE 802.3bs, 200 Gb/s and 400 Gb/s Ethernet, for 400GBASE-SR16, specifying 100 m reach. The same references were added to draft standard IEEE 802.3cd, 50 Gb/s, 100 Gb/s, and 200 Gb/s Ethernet, for 50GBASE-SR, 100GBASE-SR2 and 200GBASE-SR4, specifying 100 m reach, in November 2016. The references in both the 802.3bs standard and the 802.3cd draft standard were updated to OM5 and fiber model A1a.4 per updated IEC and ISO/IEC standards in 2Q17. 802.3bs was completed in 2017 and 802.3cd is scheduled to be completed in 2018. The supported link distance for OM5 will be equal to OM4 reach, as all these links operate only at 850 nm.

3.4.2 IEEE 802.3 Next-generation 200 Gb/s and 400 Gb/s MMF PHYs Study Group Consensus was developed in IEEE 802.3 New Ethernet Applications (NEA) Ad Hoc meetings to request a Study Group (SG) for next-generation 200 Gb/s and 400 Gb/s MMF Physical

Layer Devices (PHYs). A Call For Interest (CFI), the first step in creating an IEEE 802.3 Ethernet standard, for “Next-gen 200 and 400 Gb/s PHYs over Fewer MMF Pairs” [14] was generated and presented in the November, 2017 802.3 Plenary meeting. The phrase “fewer fiber pairs” recognized that Ethernet standards are either in progress now for 200GBASE-SR4 over 4 pairs of MMF or completed for 400GBASE-SR16 over 16 pairs of MMF. 55 individuals from 38 companies were supporters for the CFI, including cloud and enterprise end-users. At the CFI, 56 individuals from 24 companies indicated participation in this project. During November 2017 Plenary, 802.3 Working Group (WG) Voters approved the motion, with almost unanimous support, to form the Next-generation 200 Gb/s and 400 Gb/s MMF PHYs Study Group. The principal technology enablers for 200 Gb/s and 400 Gb/s PMDs on fewer MMF pairs are: 1. Multiple wavelengths on MMF – first commercially introduced in 2013 as 40G-BiDi and 2. VCSELs supporting 50 Gb/s, Pulse Amplitude Modulation, PAM4 signaling.

IEEE 802.3cd selected 26.5625 GBaud signaling with PAM4 modulation to implement 50 Gb/s to meet its objective to “Define single-lane 50 Gb/s PHY for operation over MMF with lengths up to at least 100 m.” [14]. Both FIT and Finisar have demonstrated technical feasibility of supporting 50 Gb/s PAM4 signaling and sample quantity products are now available.

In terms of economic feasibility, adding wavelengths and PAM4 modulation allows MMF modules to retain their historic cost and power advantages over SMF modules: Some major cost, power and operational advantages for VCSEL-MMF links are: relaxed alignment tolerances (~10x) for laser, mux/demux, and connectors; MMF connectors more resilient to dirt; lower drive currents (5-10mA vs. 50-60mA) and on-wafer testing. These core benefits will persist when comparing 400Gb/s technologies for short reach along with: gearbox function is needed to covert native 50Gb/s PAM-4 to 100Gb/s PAM4 for SM 400G-DR4 and 400G-FR4; laser Relative Intensity Noise (RIN) reduction for PAM4 is as, or more difficult for Distributed Feedback (DFB) lasers as for VCSELs; packaging for 1310 nm sources at 100 Gb/s per lane PAM4 has required significant development. The IEEE 802.3 Next-generation 200 Gb/s and 400 Gb/s MMF PHYs Study Group (SG) reviewed contributions supporting technical feasibility, economic feasibility, broad market potential, distinct identity and objectives for: 200 Gb/s over 1 pair of MMF with lengths up to at least 100 m - included multiple contributions and support from 19 large enterprise and multi-tenant data center expert end users affiliated with 15 different global companies across wide range of industry segments and regions, and from other industry experts; 400 Gb/s over 4 pairs of MMF with lengths up to at least 100 m - included multiple contributions and key contribution/support from cloud/hyperscale industry expert affiliated with one of the largest Chinese hyperscale cloud companies and from other industry experts; and 400 Gb/s over 8 pairs of MMF - included multiple contributions and key contribution/support from cloud/hyperscale industry expert affiliated with one of the largest US hyperscale cloud companies and from other industry experts. [25] [26].

Those experts and end-users supporting a 200 Gb/s over 1 pair of MMF objective highlighted the following key benefits/need: “200G uplink solutions are needed to support server data rate upgrade path; defining a 1-pair (duplex) solution for 200G will lower cost; enables lower cost upgrades for customers with 1-pair uplink cabling infrastructure; removes upgrade hurdles by

eliminating the need for cabling change; retains existing operational paradigm, lowering operational costs.” [27] While a 63% majority of the SG voters supported the 200 Gb/s over 1 pair MMF objective, however, the required super-majority of 75% approval was not met, so this objective was not adopted by SG.

The 400 Gb/s over 4 pairs of MMF objective was approved by both SG and WG. Key contribution highlighted the following key benefits/need: major Chinese cloud/hyperscale data center company currently utilizes 100G-SR4, finding it more cost-effective than single mode alternatives based on either PSM4 or CWDM4 optics [28] (customer and architecture specific) and expects to migrate to 400 GbE by 2019; retaining the well-established 4-pair or SR4 quad paradigm provides the most cost-effective and needed upgrade path. Large enterprise data centers have similar needs. [28] [29] [30].

The 400 Gb/s over 8 pairs of MMF objective was also approved by both SG and WG. Key contribution, authored by industry expert affiliated with major US cloud/hyperscale company, highlighted the following benefits/need: “Flexibility: 400G-SR8 offers flexibility of fiber shuffling with 50G/100G/200G configurations. It also supports breakout at different speeds for various applications: compute, storage, flash, GPU, and TPU” [13] and “High density: 400G-SR8 OSFP/QSFP DD transceiver can be used as 400GBASE-SR8, 2x200GBASE-SR4, 4x100GBASE-SR2, 8x50GBASE-SR.” [13]. “An 8-pair medium gives access to the SerDes rate for 400GAUI-8 and future 800GAUI-8; octal paradigm and form factors utilized.” [30]

The summary of the key adopted objectives are: 1. 400G over 4 pairs MMF up to at least 100 m 2. 400G over 8 pairs MMF up to at least 100 m. There were additional contributions for 400G-SR1.8, 400 Gb/s over 1 pair of MMF [31], and for 400G-SR4.1 based on 100 Gb/s PAM4 VCSELs [32]. Both these contributions were seen as too early from a technical feasibility perspective for the Next-generation 200 Gb/s and 400 Gb/s MMF PHYs Study Group, but they show continued possible future paths for VCSEL/MMF links.

The SG has generated and approved the Project Authorization Request (PAR), Criteria for Standards Development (CSD) and Objectives required for the standard to move to Task Force (TF), the project phase where the standards document is generated and finalized. The 802.3 Working Group has also approved the PAR, CSD and Objectives. Further required approvals were provided by IEEE hierarchy for PAR and CSD, successfully completing the SG phase in May, 2018.

3.4.3 IEEE P802.3cm 400 Gb/s over Multimode Fiber Task Force IEEE P802.3cm Task Force has had two meetings. A 400GBASE-SR8 Baseline was adopted in May, 2018 [33], and the 400GBASE-SR4.2 Baseline was adopted in July, 2018 [34]. Project timeline was adopted in September 2018 and updated in January 2019. Draft D1.0 of standard was authorized in September 2018, Task Force is currently working on completing Draft D1.2, expects authorization of Draft 2.0 by Task Force and Working Group in March 2019, and the final standard will be completed in December 2019.

The two 400 Gb/s MMF Physical Medium Dependent (PMDs) being standardized are shown in Figure 16. 400GBASE-SR4.2 meets the 4-pair objective and uses 2 BiDi wavelengths per fiber. 400GBASE-SR8 meets the 8-pair objective, does not use WDM, but requires double the number of fibers (relative to SR4.2). Both

are 400GAUI-8 electrical and are 8x50G electrical to 8x50G optical.

Figure 16: 400G MMF being standardized in IEEE

P802.3cm 400GBASE-SR4.2 is the first IEEE 802.3 standard to use wavelength division multiplexing (WDM) for MMF PMD and is a two wavelength solution. It is the first standards based application that will take advantage of OM5 fiber’s higher bandwidth at longer wavelenghts. The adopted baseline has reach objectives of 70 m OM3, 100 m OM4 and 150 m OM5, and utilizes a standard 12-fiber MPO connector interface, with 8 active fibers. There is currently a proposal for Optical Media Dependent Interface (MDI) lane assignment under consideration. 400G-SR4.2 provides a low-cost point-to-point link, while also allowing a 4x100G breakout, even though 100G-SR1.2 is not an IEEE standard and is not in current scope of 802.3cm project. A 4x4 fiber shuffle allows a 32-port 400G switch to be used as a 128-port 100G switch (see Figure 17)

Figure 17: 400GBASE-SR4.2 point-to-point, breakout and

shuffle modes

400G-SR8 is an eight fiber-pair solution and has reach objectives of 70 m on OM3 and 100 m using OM4/OM5. Both single row MPO-16 and 24f MPO (MPO-12 two-row) were chosen as MDI options (see Figure 18) [35]

Figure 18: 400G-SR8 adopted connector options for MDI

[35] 400GBASE-SR8 optics offer maximum flexibility for shuffle applications. Figures 19 and 20 show point-to-point, 2x2 fiber

shuffle, 4x4 fiber shuffle and 8x50G breakout capabilities and applications.

Figure 20: 400GBASE-SR8 point-to-point and shuffle

modes

Figure 20: 400GBASE-SR8 breakout

3.5 Next Generation: 100 Gb/s VCSELs It is expected that Top of Rack (TOR) switches will give way to Middle of Row (MOR) switches as server counts in racks decrease, due to increasing power dissipation per server; and that server speeds will move to 100G (see Figure 21) [36]. VCSEL/MMF-based links over MMF will evolve to 100 Gb/s PAM4 VCSELs in the approximately 2021 timeframe, initially supporting short-reach server interconnects in an 800GBASE-SR8 type application offering 8-way breakout to 30 m, but will support 100 m switch links longer term.

Figure 21: Evolution from TOR switch architecture to

MOR switch architecture [36] 100G VCSELs will provide the next step in the evolution of and contiuned relevance of VCSEL-MMF links.

3.6 Possible Next Generation MM Links Current MM technology allows speeds of up to 200 Gb/s in a single fiber pair by combining WDM and advanced signal modulation. Parallel connectivity utilizing 4 fiber pair SDM can increase the maximum link speed to 800 Gb/s. Further increases

Booth,Issenhuth,Microsoft

in transceiver baud rates may double the above numbers to 400 and 1600 Gb/s for duplex and 4 pair links, respectively, but in order to preserve, or extend, the current 100 meter standard reach limit, a joint effort on behalf of fiber, cable and transceiver makers will be required. On the fiber side, the dominant fiber parameter that limits reach is high chromatic dispersion (CD): around 100 ps/nm/km at 850 nm. Chromatic dispersion decreases at wavelengths beyond 850 nm in standard high silica fibers, dropping to around 65 and 34 ps/nm/km at 950 and 1060 nm, respectively. Together with lower fiber attenuation and other decreasing link penalties, this presents a clear opportunity to further improve link performance and reach at wavelengths beyond 953 nm [37]. For example, extending the fiber operational window to 1060 nm accommodates another 4 WDM channels beyond 953 nm, while leveraging and improving the benefits of the current mature and reliable VCSEL technology. At the same time, improvements to transceiver rise/fall times, spectral laser linewidth and adjustments to worst-case wavelengths may be needed to fully take advantage of longer wavelengths and to preserve the link reach. Ultimately, the successful technology will have to leverage both improved fiber and transceiver properties in order to minimize total link costs in a very cost-sensitive short reach datacenter environment, maximize link performance and to preserve the MM link paradigm in the future. 3.7 SWDM4 MSA, 100G-BiDi & 400G-BiDi MSA The SWDM MSA developed and released specifications for 40G SWDM4 and 100G SWDM4 on March, 2017. Commercial 40GSWDM4 and 100G SWDM4 products were available by the end of 2017. The key value proposition of SWDM4, which utilizes 4 wavelengths on 30nm spacing of 850nm, 880nm, 910nm and 940nm, is to utilize much lower cost duplex fiber cabling and connectivity instead of parallel fiber cabling and connectivity.

100G BiDi uses 2 wavelengths, 857 & 908nm, to also extend the value of duplex cabling and connectivity infrastructure. Commercial 100G BiDi products became available at the beginning of 2018.

The 400G BiDi MSA was announced in July, 2018 to define “optical data link specifications and promoting adoption of interoperable 100 Gb/s and 400 Gb/s optical transceivers for 100 meter link distance based on a dual wavelength bidirectional transmission technology in multi-mode optical fiber (MMF).” The reach objectives are 70 m OM3, 100 m OM4 and 150 m OM5. Specifications are currently in development and will be available later this year.

The supported reaches for OM3, OM4 and OM5 for 100G-SWDM4, 100G-BiDi, 400G-BiDi, along with MMF PMDs from 802.3bs, 802.3cd and 802.3cm are shown in Table 1 below highlighting the extended reach benefit and future-proofing value proposition of OM5:

Table 1. Evolution of VCSEL-based links over MMF

4. Conclusions Key hyperscale cloud data center companies in North America and China, and large enterprise data center customers have continued to leverage the cost savings, power and usability advantages of MMF SR4 optics as they migrate from 40GbE to 100GbE as evidenced by the robust market for 100GBASE-SR4 in 2016-2017. A 50 Gb/s ecosystem to support 400G switching is under development, with enabling technologies that include 50 Gb/s PAM4 signaling and the use of multiple wavelengths on MMF.

Two and four wavelength BiDi and SWDM4 duplex MMF transceiver solutions are available for 40GbE and 100GbE. The recently announced 400G-BiDi MSA continues the WDM MMF solutions industry trend. IEEE P802.3cm Task Force has adopted a baseline for 400GBASE-SR4.2 and is rapidly progressing to create that standard – IEEE’s first WDM standard for MMF links. These multiple wavelength solutions will allow higher data rates over fewer fiber pairs and are supported by both new and installed base of OM3/OM4 fiber. However the maximum benefit is achieved with the new industry standard OM5 fiber that provides enhanced reaches, e.g. 50% more reach at 100GbE, increasing the value proposition and future-proofing of extended reach MMF solutions for >40GbE (possible future 200 GbE and 400 GbE) duplex and >400 GbE (possible future 800GbE) four pair links. 802.3cm Task Force is also developing the 400GBASE-SR8 PMD standard, which provides maximum breakout and shuffle capabilities.

Multiple high profile hyperscale cloud companies have 400G-SR4.2 and 400G-SR8 on their roadmaps. Large enterprise data centers that are being transformed for cloud computing are forecast to continue to grow at healthy rates, providing the traditional advantages of cost, power and usability for MMF/SR/SR4 Ethernet, with OM5 having the added benefit of taking full advantage of SWDM and BiDi multiple wavelength technologies. Enterprise DCs with Fibre Channel or FCoE or FC-NVMe storage are ideally suited to continue to benefit from OM3/OM4/OM5-VCSEL solutions. In addition, high-growth edge data centers benefit from the short reach, cost, power and usability advantages of OM3/OM4/OM5-VCSEL solutions

In summary, key hyperscale cloud data centers and enterprise data centers continue to be a strong demand base for short reach, cost effective multimode solutions. In addition, close to end customer edge, off-premises co-location and telecom/CO data centers provide new and growing addressable markets for the cost/power advantages of short reach <100 m OM3/OM4/OM5-VCSEL links. Industry standard OM5 solutions are optimized for multiple wavelength SWDM and BiDi transceivers to provide enhanced reach and value. IEEE P802.3cm Task Force is standardizing 400 Gb/s MMF PMDs that leverage and address these market and technology trends, underscoring the industry commitment and continued end-user demand for MMF/VCSEL solutions, and to extend the future of MMF.

5. References [1] Finisar. http://myvcsel.com/key-advantages-of-vcsel-

technology/ [2] Ethernet Alliance. The 2018 Ethernet Roadmap

https://ethernetalliance.org/the-2018-ethernet-roadmap/ [3] FCIA, The Fibre Channel Roadmap

https://fibrechannel.org/roadmap/ [4] Dell’Oro May 2017

[5] FCIA, Fibre Channel Solution Guide 2017 http://fibrechannel.org/wp-content/uploads/2017/07/FCIA_SolutionsGuide2017_WEB.pdf

[6] Cisco VNI Global Fixed and Mobile Internet Traffic Forecasts. 7th annual Cisco Global Cloud Index (GCI) forecast. Global data center and cloud computing trends (2016 – 2021). February 2017

[7] National Institute of Standards and Technology “The NIST Definition of Cloud” Special Publication 800-145. September, 2011

[8] Cisco blog “AWS or Private Cloud or both, what’s your strategy?” November 2016 https://blogs.cisco.com/datacenter/aws-or-private-cloud-or-both-whats-your-strategy

[9] “Cisco Global Cloud Index: Forecast and Methodology, 2016–2021 White Paper” Document ID: 1513879861264127. February 2018

[10] C. Xie, Alibaba OIF Q4 2017 Shanghai [11] C. Gang, Baidu, 2018 Optinet, Shanghai [12] X. Zhou, Google, OFC 2018, San Diego [13] Zuowei Shen, Google “400G-SR8 Broad Applications for

Datacenters” IEEE 802.3 NGMMF SG, March 2018 http://www.ieee802.org/3/NGMMF/public/Mar18/shen_NGMMF_01a_mar18.pdf

[14] Robert Lingle, Jr., OFS “Next-gen 200 and 400 Gb/s PHYs over Fewer MMF Fiber Pairs than in Existing Ethernet Projects and Standards” CFI Consensus Presentation, IEEE 802.3 Plenary, November 7, 2017.http://www.ieee802.org/3/cfi/1117_4/CFI_04_1117.pdf

[15] Ryohei Urata, Hong Liu, Xiang Zhou, and Amin Vahdat, “Datacenter Interconnect and Networking: from Evolution to Holistic Revolution” ryohei@google.com OFC 2017

[16] Dell’Oro Ethernet Switch Data Center 5 Forecast. January 2018. Fromhttp://www.ieee802.org/3/NGMMF/public/Mar18/kolesar_NGMMF_01b_mar18.pdf (page 15)

[17] “The virtualized and converged central office/cable headend” White Paper, CommScope, February 2018

[18] CORD (Central Office Re-architected as a Datacenter) platform https://www.opennetworking.org/cord/

[19] “Top Factors Driving the Global Edge Data Center Market – Technavio” Business Wire, March 2018

[20] “Future Ready: Top Data Analytics Platforms to Leverage Edge Computing” TechGenix, September 2017

[21] LightCounting Market Research, August 2018 [22] LightCounting Market Research, August 2018 [23] Worldwide Multimode Fiber Demand, CRU, Feb 2018 [24] LightCounting Market Research, August 2018 [25] Contributed Presentations for IEEE 802.3 Next-

generation 200 Gb/s and 400 Gb/s MMF PHYs Study Group January 22-23, 2018, IEEE 802.3 Interim Meeting, Geneva, Switzerland http://www.ieee802.org/3/NGMMF/public/Jan18/

[26] Contributed Presentations for IEEE 802.3 Next-generation 200 Gb/s and 400 Gb/s MMF PHYs Study Group March 6-7 2018, IEEE 802.3 Plenary Meeting, Roseland, IL, US http://www.ieee802.org/3/NGMMF/public/Mar18/

[27] Carl Rumbolo, Wells Fargo and Jim Young, CommScope, “In Support of 200G over 1 pair MMF Objective: Broad Market Potential (BMP) & Economic Feasibility (EF)” IEEE 802.3 NGMMF SG, March 2018 http://www.ieee802.org/3/NGMMF/public/Mar18/young_NGMMF_01a_mar18.pdf

[28] Stephen Hardy, “IEEE Study Group explores future multimode fiber roles” Lightwave, February 16, 2018

[29] Xie, Alibaba, Piehler, Dell EMC, Lingle, Jr., OFS “Broad market potential, economic feasibility, and distinct identity for an 400GBASE-SR4.2 objective” IEEE 802.3 NGMMF SG, January 2018 http://www.ieee802.org/3/NGMMF/public/Jan18/lingle_ngmmf_03_jan18.pdf

[30] Robert Lingle, Jr., OFS “In support of Broad Market Potential (BMP) & Distinct Identity (DI) for both 4-pair & 8-pair MMF objectives at 400G” IEEE 802.3 NGMMF SG, March 2018 http://www.ieee802.org/3/NGMMF/public/Mar18/lingle_NGMMF_01_mar18.pdf

[31] Vipul Bhatt , Finisar “The Need for 400G Duplex MMF Objective” IEEE 802.3 NGMMF SG, January 2018 http://www.ieee802.org/3/NGMMF/public/Jan18/bhatt_NGMMF_01_jan18.pdf

[32] Ali Ghiasi, Ghiasi Quantum “The Need for 100Gb/s/lane MMF PMDs” IEEE 802.3 NGMMF SG, January 2018 http://www.ieee802.org/3/NGMMF/public/Jan18/ghiasi_NGMMF_01_jan18.pdf

[33] “Straw Polls & Technical Motions” Motion #4, IEEE P802.3cm TF, May 2018 http://www.ieee802.org/3/cm/public/May18/straw_polls_and_technical_motions_3cm_0518.pdf

[34] “Straw Polls & Motions” Motion #3, IEEE P802.3cm TF, July 2018 http://www.ieee802.org/3/cm/public/July18/straw_polls_and_motions_3cm_01_0718.pdf

[35] Paul Kolesar, CommScope “400GBASE-SR8 MDI Choices” IEEE P802.3cm TF, May 2018 http://www.ieee802.org/3/cm/public/May18/kolesar_3cm_01_0518.pdf

[36] Tom Issenhuth and Brad Booth, Microsoft “100G-DR Use Cases & End User Perspective” IEEE P802.3cd TF, September 2016 http://www.ieee802.org/3/cd/public/Sept16/issenhuth_3cd_01a_0916.pdf

[37] T. Kise et al. “Development of 1060nm 25-Gb/s VCSEL and demonstration of 300m and 500m system reach using MMFs and link optimized for 1060nm,” in Proceedings of the Optical Fiber Communication Conference, paper Th4G.3, 2014

6. Pictures of Authors

Robert Lingle, Jr. is Director of Systems & Technology Strategy at OFS in Norcross, GA, as well as Adjunct Professor of Electrical and Computer Engineering at the Georgia Institute of Technology. He has a research background in short pulse lasers and their application to fundamental processes in liquids and interfaces, with a Ph.D. in physics from LSU and a postdoc in surface physics at UC Berkeley. At Bell Labs and now OFS, he first worked in sol-gel materials chemistry and then managed the development and commercialization of many new optical fiber products. His current research is in the confluence of optical and electronic methods for mitigating impairments in optical communications systems.

John Kamino is a Senior Manager – Product Management for OFS (www.ofsoptics.com). A 27-year veteran of the company, he has also held positions in optical connectivity product management, offer management, product marketing, sales, and engineering. He has a B.S. in Chemical Engineering from the University of Nebraska-Lincoln, and a M.B.A. from Mercer University.

Durgesh S. Vaidya is Director, Global Fiber and Connectivity R&D, responsible for Optical Fiber and Optical Connectivity Solutions innovation in Ocean, Long Haul Terrestrial, Metro/Access, Datacenter, High Performance Computing, and Fiber to The Home, and Mobile Backhaul applications. At Bell Labs and now OFS, he leads global product and process development projects, commercialization of new products, and improvements in materials and fabrication methods for cost-effective scale-up of optical system and network performance. He holds a Bachelor’s in Chemical Engineering from the Indian Institute of Technology, Mumbai, India, and a Ph.D. in Chemical Engineering from SUNY at Buffalo, NY. He serves as a Symposium Committee Member for IWCS.

David Braganza David Braganza is Technical Manager for Fiber R&D at OFS. He received the B.E. degree in electrical engineering from Pune University, Pune, India, and the M.S. and Ph.D. in Electrical Engineering from Clemson University, Clemson S.C., USA. In 2007 he joined the optical fiber R&D group at OFS where his early work focused on the development of novel processes for optical fiber manufacturing. Dr. Braganza is a senior member of IEEE.

Roman Shubochkin is a Senior R&D Engineer and a Distinguished Member of Technical Staff at OFS. He received BS/MS in Optical Engineering from Moscow Power Engineering Institute and MS/PhD in Electrical Engineering from Brown University, Providence RI. He worked as a research fellow at the Fiber Optics and Solid State Physics Divisions of the General Physics Institute in Moscow, Russia, as well as a senior research associate at Boston University. His research background includes design and fabrication of specialty optical fibers, research on fiber lasers, fiber optic sensors, pulsed solid state lasers, and rare-earth doped glasses. He joined the Optical Fiber R&D group at OFS in 2012 and works on the design, fabrication, and testing of novel transmission optical fibers. Dr. Shubochkin serves as a Vice Chair for the TR42.12 TIA Subcommittee on Optical Fibers and Cables and represents OFS at the Fibre Channel INCITS/T11 Technical Committee. Dr. Shubochkin is a member of OSA and Sigma Xi.

G. Mabud Choudhury is Standards Manager for OFS. He represents OFS in IEEE 802.3 Ethernet, INCITS T11 Fibre Channel & IEC standards. Previously, he was R&D Engineer and Manager for product development, systems engineering and standards of copper, fiber and intelligent cabling infrastructure products/solutions for AT&T Bell Labs, Lucent, Avaya, and CommScope. He has an MS degree in Mechanical Engineering from Massachusetts Institute of Technology, and a BS degree in Mechanical Engineering and in Chemistry from Duke University. He holds 32 US patents and has authored several papers, webinars and IEEE/Fibre Channel/TIA standards contributions.

Yi Sun has a BS/MS in Astronomy from Peking University, a MS in EE and a Ph.D. in Physics from Northwestern University. Dr. Yi Sun joined Bell Laboratories in 2001 and currently is a distinguished member of technical staff in OFS Fitel, LLC in USA. Dr. Sun's research work covers various novel optical fibers and their applications in Datacom and SDM. Dr. Sun is the holder of 12 patents and has authored more than 100 publications. Dr. Sun receives several outstanding contribution awards by OFS and the 2011 Spirit Endeavor Award by TAG. Dr. Yi Sun is a senior member of IEEE.