White Paper

Moving to 100G & Beyond

Prepared by Sterling Perrin Senior Analyst, Heavy Reading www.heavyreading.com

on behalf of

www.zte.com.cn

December 2012

Introduction 100G transport is a great optical industry success story. From the first calls to actions made by network operators some six years ago, the technology has progressed steadily through hero experiments, standardization, field trials and now, commercial adoption. As of the third quarter of 2012, Heavy Reading tallied more than 170 commercial 100G wins globally, up from just a handful of such wins in 2010. As 100G deployments expand globally, industry participants have set their sights on the next phase of transport beyond 100G. This white paper assesses both the state of commercial 100G today and industry progress toward the next transport rate beyond 100G. On the 100G front, the paper details the key applications drivers for 100G as well as systems requirements. The paper describes the enabling technologies that have made 100G a success and presents Heavy Reading's current 100G forecast through 2015. Looking beyond 100G, the paper provides the state of standardization efforts from both the IEEE and the ITU-T. It then discusses the major technology innovations required to get to the next bit rate, including super channels, flexible spectrum, and advanced modulation. The paper concludes with three operator case studies of both 100G deployments and beyond 100G trials.

100G Market Assessment Applications Drivers The continued growth in IP traffic is the primary driver for 100G transport. Five years ago, carrier IPTV build-outs and the rise of over-the-top video (e.g., YouTube and Netflix) placed wireline video as the key single application but in recent years other applications have emerged as major drivers. On the wireline side, cloud application delivery is commonly cited by operators as the primary driver of IP growth and, therefore, become commonly associated with 100G. In addition, the migration from 3G to 4G LTE mobile technology places an unprecedented burden on metro and long-haul networks that carry this high-speed access traffic. Network operators commonly view 4G builds as the trigger for a major network upgrade and 100G is being added as a requirement to that technology upgrade list. Initial interest in 100G centered on 100GE router interconnect. This particular application is definitely a part of early 100G rollouts, but 100G transponders are greatly outnumbered by 10x10G muxponders, in which multiple 10G links are aggregated for transport at 100G. This means that the 100G trend is much broader than high-speed router interconnect. Heavy Reading surveys have shown that, in addition to router interconnect, network capacity exhaust and better economics compared to 10G are the main 100G deployment drivers. Given the long incumbency of 10G in the long-haul network, coupled with the slow ramp of 40G, this broad appeal of 100G should not be surprising. 10G became the dominant rate for long-haul networks in 1999 and remains the dominant rate – in terms of both the number of interfaces and the total capacity shipped – today. Meanwhile, during this 12-year period, IP traffic has continued to increase at triple-digit and high double-digit rates annually. Some operators have migrated to 40G, but a fragmented supply chain, high prices and a late start all contributed to limited success for this rate.

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

2

With 100G commercially available and with a strong supply chain, operators have settled on 100G as the next-gen transport rate for their long-haul networks. There is significant opportunity for 100G in regional and metro networks as well, Heavy Reading believes.

Key System Requirements The fundamental mission of optical transport is to reduce the cost per bit for transport. As such, cost reduction is critical to the near- and long-term success of 100G. Heavy Reading surveys have shown that lowest overall system cost is the highest-priority requirement in evaluating 100G systems, followed closely by the ability to reuse existing infrastructure, then by transponder costs. These requirements are all related and relate back to the fundamental mission of optical transport. Much of the 100G pricing public discussion centers on transponder costs, but our operator surveys show that operators have a more sophisticated view of pricing. By using their existing network infrastructure – particularly the existing line systems – operators can dramatically lower their overall costs for introducing 100G. On the other hand, a 100G system that promises relatively low transponder costs but that also requires a major line-side upgrade may not prove in economically, and will likely be rejected for a system that carries a higher transponder cost but that works with the existing infrastructure. As indicated in our surveys, transponder costs are still important as they relate to: 1.

The overall cost of a system, and

2.

The decision whether to continue to deploy 10G wavelengths or migrate to 100G wavelengths.

We note that 40G struggled to gain traction in the market because it could not reach cost parity with 10G – meaning it was less expensive to deploy four 10G cards than one 40G card. Operators that we have interviewed have called for 100G transponder pricing at 6x 10G pricing. This is an ambitious target that has not been reached, but, significantly, it has not stopped early deployments of 100G. Suppliers have been able to reach cost parity with 10G – meaning it costs the same amount or less to deploy one 100G card than to deploy ten 10G cards – and this has been enough for operators to make the migration. As volumes ramp and technology continues to mature, we expect the 6x 10G pricing threshold will ultimately be reached.

100G Forecasts Since its start in 2009, the commercial 100G market has grown tremendously and is well on its way to becoming a mainstream backbone technology, while also expanding into regional as well as metro core networks. As of the third quarter of 2012, Heavy Reading counts more than 170 commercial 100G wins to date. This is a tremendous change from the single-digit number of wins that were tallied in 2010. Significantly, these deployments are occurring in every major region of the world and include some of the world's leading Tier 1 operators, including Verizon, Deutsche Telekom, China Telecom, Mobily and others. Figure 1 shows Heavy Reading's forecast for long-haul DWDM capacity shipped through 2015. Starting from virtually nothing in 2010, 100G is forecast to rise to the largest share of long-haul capacity shipped by 2015, exceeding both the incum-

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

3

bent 10G technology and the high-speed challenger 40G. Heavy Reading has long predicted that the advent of commercial 100G would lead to the decline of the 40G opportunity. Based on the 100G uptake we have seen in 2012 and operator plans for 2013, we believe that the rise of 100G, at the expense of 40G, could be more rapid than previously predicted. Figure 1: Worldwide Long-Haul DWDM Share of Line-Side Capacity by Speed

Source: Heavy Reading, 2011

100G Enabling Technologies ASIC Chip Design The advent of coherent detection in DWDM transmission heralded a major shift in optical transport: for the first time, the electronics portion of the line card became as important as the optical portion – in terms of its impacts on performance as well as cost. This means that starting at 100G, and for any rates beyond 100G, electronics and optics will play equal roles in driving and inhibiting line rate advancements. The general trend across semiconductor processes is the continuous move to smaller geometries and the increase in integration and reduction in size and cost that this enables. In semiconductors, the metric used is physical gate length measured in nanometers (nm). As an example of the benefits gained in moving from one generation to the next, a 40nm process more than doubles gate density over 65nm and provides 40 percent better performance and a 35 percent reduction in power. One key recent advancement in electronics for 100G DWDM transport is the commercial introduction of 40nm silicon ASICs. For 100G, these 40nm ASICs are being used for functions including soft-decision forward error correction (FEC), digital signal processors used for coherent detection, and ultra-fast analog-digital and digital-analog conversions. In addition, for beyond 100G, these chips meet processing requirements for sophisticated modulation schemes, such as 16QAM. We discuss coherent detection and soft-decision FEC in more detail below.

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

4

Coherent Detection Coherent detection is a key technology enabler for 100G. Like QPSK modulation, it was borrowed from radio communications and applied to optical communications. While coherent detection is useful for 40G transport, it is critical for mitigating transmission impairments that become worse as transmission distance increases in long-haul 100G transport. A coherent receiver is able to access the amplitude, phase and polarization of the incoming signal in the electrical domain. DSPs then compensate for chromatic dispersion and polarization mode dispersion impairments – again, all handled electronically. The primary benefit of coherent detection with DSPs in 100G transmission is the ability to maintain both high performance, including 2,000km+ transmission distances, and high spectral efficiency. Another benefit of using electronic dispersion compensation is that it eliminates the need for costly dispersion-compensating fiber spools, along with the associated amplifiers, that would otherwise be required for long-distance transmission. It is important to note that direct detection can be used at 100G speeds, but transmission distances are greatly reduced due to the effects of coherent detection and polarization mode dispersion, as noted above. Some vendors have developed single wavelength 100G transport technology using direct detection and targeting application in the 40-600km range. Yet, as coherent 100G costs have come down dramatically over the past year, industry momentum is behind coherent 100G, even for these shorter-distance applications. Heavy Reading believes that we are moving into the digital coherent era, and that 100G is the first rate that will truly take advantage of these channel-boosting innovations. A key point is that coherent detection will be used at rates beyond 100G well into the future.

Soft-Decision FEC FEC has been used in DWDM transmission for more than a decade to lower bit error rates and, as a result, squeeze longer reach out of DWDM systems. The form of FEC employed historically 10G and 40G networks has been hard-decision FEC. With hard-decision FEC, a firm decision is made as to whether a bit of information is a 1 or a 0, and this information is fed to the FEC decoder. Soft-decision FEC, by contrast, is a more complex technique that uses additional bits of data. The soft-decision FEC decoder delivers a 1 or a 0 decision, just like a hard-decision decoder, but it also provides a probability factor in its decision indicating how likely the decoded bit is actually a 1 or a 0. As a result of the more detailed information delivered to the decoder, soft-decision FEC offers improved performance/greater distances compared to hard-decision FEC. Based on our discussions, suppliers supporting soft-decision FEC at 100G cite a 1-2dB gain compared to hard-decision FEC. The 1-2dB gain delivers a 20-40 percent increase in 100G system reach, suppliers report. One tradeoff is that soft-decision FEC increases transmission overhead, increasing overhead bytes from 7 percent using earlier generation FEC to 15-20 percent overhead using soft-decision algorithms. This means that a 120-128Gbit/s bit rate is required in order to transmit 100 Gbit/s of data. Another tradeoff is that soft-decision FEC greatly increases processing complexity, which is why DWDM systems suppliers historically have favored various hard-

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

5

decision techniques. Recent advances in semiconductors, however, have made soft-decision FEC both technically and economically feasible for DWDM systems. As a result, beginning at 100G transmission, we are seeing a migration to softdecision FEC.

OTN Framing & Switching OTN has rapidly risen as the bandwidth management and grooming technology of choice for the next generation of core transport networks. Several trends have converged to make OTN the consensus choice for core networks. Operators are pushing transport down to the lowest feasible layer of the OSI stack in order to lower the cost per bit for transport. While all-optical transport is the least costly form for wavelength-level traffic (at Layer 0), it is not efficient for transporting subwavelength level traffic. OTN, by contrast, is able to groom traffic to fill wavelengths and is less costly than performing these functions at Layers 2 or 3. In addition, as a Layer 1 TDM-based technology, OTN is widely accepted by operators as an ideal technology for transporting and grooming legacy Sonet/ SDH traffic. While Sonet/SDH is not growing, a large installed base of Sonet/SDH remains, particularly in incumbent operator networks, and OTN is suitable for this traffic type. OTN differs from Sonet/SDH in its ability to handle IP/packet traffic. OTN was created as a universal transport protocol to handle TDM traffic and packet traffic, much more efficiently than legacy Sonet/SDH. Standards advances such as ODU0 (for Gigabit Ethernet) and ODUflex have further adapted OTN for the packet transport and switching role. Heavy Reading views OTN and 100G are tightly coupled trends. The last bit rate defined for Sonet/SDH was 40G. Starting with 100G, and for any transport rates beyond 100G, OTN is the framing protocol. Specifically for 100G transport, the standard is OTU-4; and the next transport rate beyond 100G will also be based on OTN. But the tight OTN/100G coupling extends beyond framing to grooming, as well. The reason is that, while operators are rapidly transitioning their core transport networks to 100G, the client side of those connections is expected to remain 10G for many years to come. This mandates an efficient grooming technology to make sure those 100G pipes are packed efficiently. Otherwise, the 100G wavelengths are lightly filled, and the efficiencies of 100G are lost. Operators widely view OTN as the best technology to fill this 100G grooming role, at least for core networks. As 100G migrates to the metro, we expect OTN will also be used there for grooming.

Evolution Beyond 100G Standards Path & Progress to Date Two standards bodies are responsible for defining the next transport rate beyond 100G. These two bodies are the IEEE, which is responsible for Ethernet applications and the ITU-T, which is responsible for OTN. The IEEE will define the Ethernet interface for the next Ethernet bit rate and the ITU-T will define the standardized OTN container for this Ethernet rate. In terms of flow, the ITU-T needs to understand the IEEE Ethernet roadmap in order to define an OTN container that efficiently transports that rate. Thus, the two groups must work together closely.

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

6

The following section discusses the IEEE and ITU-T Beyond 100G work and roadmaps in more detail. IEEE Client Side The next Ethernet bit-rate beyond 100GE is not yet defined, which is why we must speak vaguely in terms of "beyond 100G" when discussing the evolution. On the table for the client side are a 4x jump from 100GE to 400GE or a 10x jump to 1Terabit Ethernet. The IEEE 802.3 Industry Connections Higher Speed Ethernet Consensus Group was formed in August to build consensus on the next bit rate questions and is preparing a Call for Interest (CFI) on 400GE, which is slated for March 2013. Given the activities of this ad hoc group to date, 400GE appears the likely winner, but nothing is certain until the vote is completed and the official study group is formed to define the standard. Current expectations are for a completed standard in 2016 or 2017. ITU-T Line Side The ITU-T Study Group 15 began discussions of the next-gen transport rate at its Plenary Meeting in Geneva, held in September 2012. On the line side, no decisions have been made, and there are currently three options on the table: ·

400 Gbit/s

·

1 Tbit/s

·

A flexible line rate decoupled from the client rate and is built from standardized building blocks

The current goal is to complete the standard in the 2014 to 2016 timeframe, though the IEEE Ethernet rate is an important input into the ITU-T decision process and could impact the line-side timetable if there were unforeseen delays in reaching a client-side consensus. We believe that a flexible and adaptive line-side rate has strong merits and has the most momentum currently. Though no decisions have been made and discussions are ongoing, the flexible channel rate option (often called ODUCn) has been added to the living list of Beyond 100G proposal options – along with the 400G and 1Tbit/s fixed-rate options. The concept is that larger OTN containers can be created dynamically by adding or subtracting standardized building blocks, such as 100 Gbit/s or some other increment that is decided by the standard. One of the key advantages of this approach is that it mitigates the need to convene a standards body every time a new line rate is required, meaning that line rate innovation would occur more quickly than ever before. The approach also enables a de-coupling of the line side and the client side, since the line-side containers would be assembled to hold whatever client rate is presented. As a final and important point, multi sub carrier super channels fit well with a flexible line rate standard as the sub-carriers could be added and subtracted from the line to provide the right amount of capacity required for the container. The use of sub-carriers means that capacity is not wasted when channel capacity requirements are low. Idle sub-carriers are available to be used for other wavelengths. This would not be the case for ultra-high capacity, single-carrier wavelengths (such as with a single-carrier 1Tbit/s channel, for example).

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

7

Enabling Technologies Super Channels As suppliers look beyond 100G, there is increasing consensus among them that the super channel is the best solution. A super channel is an aggregation of multiple (two or more) sub-channels within a given spacing, using multiple laser sources and/or modulators. A difference between a sub-carrier based super channel and a simple aggregation of multiple channels is that sub-carriers are tightly spaced in order to increase bandwidth while also improving spectral efficiency (the measure of bit/s/Hz on the fiber). In a super channel, sub carriers are spaced at 25 GHz, or some other set granularity. One of the main drivers for using super channels is that the speed of the optics and that of the electronics do not need to match. Indeed, waiting for electronics to reach 400Gbit/s and 1Tbit/s rates to commercialize 400G and 1T transport would delay adoption by many years. As a point of reference on electronics, we are currently seeing 56G ADC converters being introduced. Thus, super channels enable more advanced optics to be combined with less advanced electronics. Flex Spectrum Flex spectrum is a next-gen ROADM function that also relates closely to Beyond 100G transport and, more specifically, super channels. Note that this functionality has also been called "gridless" and "flexible grid" – all describing the same capability. Flex spectrum is a way for operators to future-proof networks that will ultimately need to contend with transport speeds beyond 100G. For speeds beyond 100G – i.e., 400G or 1Tbit/s – more than 50 GHz of spectrum will almost certainly be required. Network operators would also like to be able to accommodate those future speeds on the same 40G and 100G ROADM networks. The proposed solution is a more granular version of the ITU grid that breaks spectrum down to 25GHz granularities. ROADM nodes supporting a flexible grid could operate at any speed that is based on increments of 25GHz spacing, such as 75GHz or 125GHz spacing, etc. See Figure 2. Figure 2: Illustration of Flexible Spectrum Architecture

Source: Verizon, 2010

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

8

Advanced Modulation The transition from simple on-off keying to advanced modulation schemes was a technology enabler for 40G transmission and 100G. For 100G transport dual polarization-QPSK has been set by the OIF as the de facto standard for long-haul transport at 100G. As the industry moves beyond 100G, to 400G and 1T transmission, advanced modulation will continue to be a major technology enabler. Suppliers are demonstrating alternative modulation schemes to DP-QPSK that are able to add additional bits per symbol and, thus, improve spectral efficiency. There have been several industry demonstrations using 16 QAM modulation to achieve 200G bit rates per carrier. Using super channels, two 200G sub-carriers can be combined for 400G data rates, or five 200G sub-carriers can be combined to achieve 1T. The physics tradeoff of using 16 QAM modulation versus QPSK is that maximum reach decreases as bits/symbol increase. For the 16 QAM demonstrations that we have heard about, distances achieved are in the regional range of 600km. Therefore, significant advances in FEC would be required in order to extend 16 QAM-based systems into long-haul and ultra-long-haul applications.

Operator Case Studies Following are three case studies aimed at showing the broad adoption of 100G today as well as the state-of-the-industry in moving beyond 100G. The first two are 100G case studies, one from China and one from Europe. The third highlights innovative work in trialing technology for beyond 100G.

China Education & Research Network (CERNET) Funded by the government and managed by the Ministry of Education, CERNET is China's first nationwide research and education network. Currently, there are more than 2,000 universities, research institutions and schools connected to CERNET, with more than 20,000 users. In November 2012, ZTE announced that it was selected by CERNET for its Phase 3 network expansion project, which includes the introduction of coherent 100G transport on the CERNET network. The new network, which began construction in 2012, will cover Beijing, Chongqing, Dalian, Guangzhou, Hangzhou, Shanghai, Shenyang, Shenzhen, Wuhan and Xi'an, with a total transmission distance exceeding 10,000km. In terms of distance covered, this is the largest 100G network in China today.

T-Mobile Austria T-Mobile Austria is a subsidiary of Deutsche Telekom that has 32 million subscribers and accounts for 29 percent of the Austrian telecom market. This year, the operator built out a 100G national backbone over its original 10G backbone. Significantly, T-Mobile Austria decided to move straight from 10G to 100G, skipping the interim step of 40G transport. Other features of the new 100G backbone include a colorless and directionless mesh ROADM network and the use of optical control software at both Layer 1 (the OTN layer) and Layer 0 (the wavelength/ ROADM layer). ZTE was selected as the 100G OTN/WDM supplier for this network. T-Mobile Austria's 100G network is the first commercial deployment of ZTE’s 100G equipment in Europe. ZTE was selected as the exclusive supplier for this network.

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

9

Deutsche Telekom Deutsche Telekom is Europe's largest telecom operator, delivering services to nearly 50 countries around the world. The operator holds a strong reputation for being a leader adopter and driver of new optical technologies, including 100G and beyond 100G. In February 2012, Deutsche Telekom and ZTE completed a field trial to test the mixed transmission of 100G, 400G and 1T over the same fiber. For the field trial, ZTE provided its commercial 100G product along with prototype versions for the 400G and the 1T transmission. Highlights of this trial included: ·

Successful data transmission of two parallel 100G, one 400G and 1T channel over a distance of 1,750km

·

Successful data transmission of two parallel 100G channels over a distance of 2,450km

·

Successful data transmission of 8x 216.4Gbit/s through 50GHz spectrum grid over 1,750km to achieve a spectral efficiency higher than 4bit/s/Hz

As noted above, Deutsche Telekom subsidiary T-Mobile Austria has commercially deployed ZTE's 100G systems.

Conclusions Within the next four years, 100G is set to become the dominant backbone technology in terms of capacity shipped – unseating longtime incumbent 10G and surpassing would-be high-speed contender 40G. OTN switching, ASIC chip design, coherent detection and soft-decision FEC are among the technology innovations that have made 100G a success. Given the continued growth in IP traffic and the five- to seven-year timeframe needed to commercialize a new line rate, Heavy Reading believes the time for R&D investment for the next transport rate is now. First, we look to the primary standards bodies, the IEEE and the ITU-T, to come to agreement internally and with each other in setting the next client-side and line-side rates. Then, we will look to the components and systems suppliers to build products that meet technical and economic requirements while allowing a smooth migration from the 100G infrastructure that is being put in operation today.

HEAVY READING | DECEMBER 2012 | WHITE PAPER | MOVING TO 100G & BEYOND

10