By Patrick McLaughlin
Several recent examples show that when the Institute of Electrical and Electronics Engineers (IEEE; www.ieee.org) 802.3 Working Group develops a specification for Ethernet, even when the specification is complete it is not the proverbial “be-all, end-all” for a given speed. On the contrary, such a specification often is just the beginning for optics developers, system designers, networking professionals and ultimate end-users; it signals the kickoff of what is likely to be a number of specifications that enable Ethernet transmission at a specific speed to some distance.
For example, among the recent initiatives within 802.3 was the formation of study groups for 25-, 50- and 100/200-Gbit/sec Ethernet transmission, in November. The IEEE 802.3 50-Gbit/sec Ethernet Over a Single Lane Study Group, and the 802.3 Next Generation 100-Gbit/sec Ethernet Study Group will jointly investigate the market requirements for 50-, 100- and 200-Gigabit Ethernet, including server-to-switch and switch-to-switch applications. The study groups will define objectives for the Ethernet specifications leveraging 50-Gbit/sec signaling technology, based on market needs. The 25-Gbit/sec Ethernet PMD(s) for Single Mode Fiber Study Group is tasked with exploring the development of new 25-Gbit/sec singlemode fiber links, as well as evaluating market requirements supporting longer-reach 25-Gbit/sec interfaces up to 10 kilometers or more. The group’s work will complement the fast-tracked efforts of the 802.3by 25-Gbit/sec Ethernet Task Force to develop a 25-GbE specification.
While publicizing and commending these initiatives, the Ethernet Alliance (www.ethernetalliance.org), commented that in making these moves IEEE 802.3 is “demonstrating Ethernet’s capacity for dynamically addressing the changing needs of its rapidly expanding marketplace.” Ethernet Alliance president Scott Kipp added, “Ethernet is beginning the standardization of a new era of speeds based on 50-Gbit/sec signaling technology. The 50-Gbit/sec lanes will enable 50-Gigabit Ethernet SFP56 modules, and 200-GbE QSFP56 modules and other corresponding technologies … The launch of these new study groups will help deliver the next generation of cost-optimized, higher-speed solutions demanded by hyperscale data centers, enterprises, cloud service providers and more. It’s proof that Ethernet will continue to be optimized for new markets.”
From MMF to SMF
During the 802.3 meeting at which the group gave the go-ahead to form these study groups, a presentation advocating the 25-GbE Over Single Mode effort elaborated somewhat on Kipp’s notion of Ethernet being optimized for different markets. Among the motivations for forging ahead were to allow markets including enterprise and metro networks to adopt 25-GbE. David Lewis of Lumentum, David Malicoat of HPE and Kohichi Tamura of Oclaro delivered the presentation, assisted by Paul Kolesar of CommScope and Peter Jones of Cisco.
The group explained that, as was the case with 1- and 10-GbE, adoption of 25-GbE over multimode fiber started earlier and ramped faster than 25-GbE over singlemode fiber. Citing market research from LightCounting, the presenters noted that 2014 was the year in which shipments of singlemode-based 1-GbE optics surpassed shipments of multimode 1-GbE optics in all form factors. Likewise, LightCounting projects 2018 to be the crossover year for 10-GbE optics, when singlemode shipments outpace multimode. In its infancy, 25-GbE is expected to follow the same early path; multimode far outpacing singlemode. However, the group pointed out that over multimode fiber, 10-GbE’s reach is 400 meters while 25-GbE maxes out at 100 meters over multimode. Therefore, the transition from multimode to singlemode for 25-GbE could be faster than what it will be for 10-GbE.
“The 25-GbE ecosystem is missing a story for >100 meters,” the presentation said. The group added that metro networks and many enterprise networks require more than 100 meters, and “leveraging 25G lane rates with 25-GbE/100-GbE just makes sense.”
The presentation referenced 25- and 100-GbE because the current 25-GbE initiative is for a “single lane,” and by providing four such parallel lanes, users can achieve 100-GbE transmission rates. In spring 2015 the IEEE finalized the 802.3bm specification, which enabled a lane-count reduction for multimode-based 100-GbE by specifying 4, 25-GbE lanes whereas the original 100-GbE (IEEE 802.ba) standard specified 10, 10-Gbit/sec multimode lanes.
This approach of using fewer, higher-speed lanes to achieve a given data rate is common within IEEE 802.3. The multiple-lane, parallel optic approach gives rise to the use of array-style, MPO connectors that facilitate multiple paths of light in each direction. As we reported recently (“Standard for 16- and 32-fiber connector interface taking shape,” September 2015), the IEEE’s 400GBase-SR16 specification will use 16 lanes of 25-Gbit/sec transmission and 16 lanes of 25-Gbit/sec reception over Om4 fiber. The Telecommunications Industry Association’s Subcommittee TR-42.13, Passive Optical Devices and Fiber Optic Metrology, is developing specifications for 16- and 32-fiber connector interfaces that will accommodate this 32-lane iteration of 400-GbE.
Also covered in that same issue (“On the fast track: WBMMF standardization,” September 2015) was the emergence of multimode-based wave-division multiplexing (WDM)-a potential alternative to serial and parallel optics designs to accommodate high speeds. Cisco brought multimode WDM to the market with its introduction of its QSFP 40G BiDi (bidirectional) transceiver in early 2014. BiDi uses duplex LC ports to enable 100 meters of 40G transmission over Om3 fiber and 125 meters over Om4 fiber.
Later in 2014 the concept of wideband multimode fiber (WBMMF) was introduced, followed shortly thereafter by the introduction of such a fiber and the initiation of a standards effort to formally specify it.
A joint task group (JTG) within TIA TR-42 has decided it will use the 32G Fibre Channel and 100GBase-SR4 link models to determine fiber parameters for the WBMMF standard. By satisfying both models, the WBMMF standard will support at least 28 Gbits/sec/wavelength to at least 100 meters, and it also will support at least 100 Gbits/sec/fiber to at least 100 meters.
Multimode-based WDM technology is an ecosystem, requiring optical fiber cabling capable of handling transmission over multiple wavelengths but also requiring the development of transceiver technology that facilitates the generation, transmission and reception of those signals.
In that vein, the Short-wave Wave Division Multiplexing Alliance (SWDM Alliance) formed in fall 2015. Upon its founding, the group commented, “Optical shortwave technology is enabled by vertical cavity surface emitting lasers [VCSELs], which are the most cost-effective lasers used in data center interconnections. VCSELS have been widely deployed at data rates up to 10 Gbits/sec, and these deployments have driven large-scale installations of duplex multimode fiber in enterprise and cloud data centers. A common technique to increase the data rate beyond 10 Gbits/sec is the use of four parallel VCSELs, each running at 10 or 25 Gbits/sec, transmitted over ribbons of parallel fiber. This technique requires eight fibers instead of two-four to transmit and four to receive. Installing such a parallel fiber can represent an expensive overhaul to the fiber plant in the data center due to the need for increased fiber capacity in the trunk and also new patch cables to the optical modules.
“By contrast, SWDM technology allows users to leverage their installed duplex multimode fiber at 40 or 100 Gbits/sec, using four VCSELs operating at different wavelengths multiplexed into a single strand of multimode fiber, thereby requiring only one transmit fiber and one receive fiber. This provides the ability to migrate from 10 to 40 or 100 Gbits/sec, while minimizing overall power dissipation and maximizing transmission distance.”
The existence of serial and parallel transmission lanes, as well as short-wave wave division multiplexing, will provide system designers with multiple architectures upon which they can build high-speed transmission systems for a number of environments.
Patrick McLaughlin is our chief editor.