TOMIZAWA LAYOUT

2/17/10

1:07 PM

Page 62

100 GIGABIT ETHERNET TRANSPORT

Optical Transport Network Evolving with 100 Gigabit Ethernet Josef Roese and Ralf-Peter Braun, Deutsche Telekom AG Masahito Tomizawa and Osamu Ishida, NTT

ABSTRACT This article overviews requirements, standardization activities, as well as enabling technologies regarding the evolution of the optical transport network with 100 Gigabit Ethernet. Joint requirements raised by the R&D activities of two international carriers are described. Technical assessment for a realistic roadmap is also described.

INTRODUCTION The need for 100 Gb transport capacity has been strengthened by the expectations of 40 Gb/100 Gb-Ethernet. Higher-speed local area network (LAN) interfaces such as 40GbE and 100GbE are required by the increasing number of end users, access opportunities, and service types. These requirements are rationalized by the sheer number of applications being envisaged such as Video (high definition with/without on-demand), latency-sensitive financial services (trade/quote/order), and network computing (its throughput doubles every 2 years). It is estimated that the server-data center will require 40GbE sooner rather than later, with several hundreds of thousands of ports by 2011; according to market forecasts, the single-mode LAN option is expected to occupy more than one quarter of all 40GbE optical interfaces. On the other hand, for the aggregation of data-centers, 100GbE will be necessary too, with a volume ramp-up around the 2013 time-frame. As LAN interfaces will be manufactured and deployed en masse, some portion of those LAN client signals will be transferred between states/prefectures/cities. According to recent experiences with dense wavelength division multiplexing (DWDM) transport, this long-reach portion will be considerable, because the traffic pattern of recent Internet Protocol (IP) services is quite different from that of ordinary telephone traffic. The traffic tends to go far (sometimes internationally) to access the server that offers the most interesting and valuable services. Based on these market needs, several carriers are requesting 100 Gb transport systems in the near future; the time frame is around 2010 for early adopters and 1–2 years later for the nextstage adopters.

S28

0163-6804/10/$25.00 © 2010 IEEE

Since international carriers have considerably different requirements, vendors have been forced to develop completely different sets of systems. International standardization can help somewhat, however agreements are often limited. Carrier-to-carrier collaboration may be effective, if common requirements can be settled. This article describes the latest information on the evolution of the optical transport network (OTN) in conjunction with 100GbE; it partly reflects the joint requirements taken from the R&D activities of Deutsche Telekom and NTT.

HIGHER SPEED 40GBE AND 100GBE ETHERNET — REQUIREMENTS AND ENABLING TECHNOLOGIES The paradigm of circuit switched networks has changed to packet switched networks due to the exponential increase in IP centric traffic created by applications like IP television (IPTV), voice over IP (VoIP), and other IP based web applications. This has increased the demand on network capacity both in the aggregation networks of the access and metro/regional networks, and in the backbone networks. The architecture of transport networks comprises multiple domains and multiple layers as shown in Fig. 1. The customers are connected via access, aggregation, metro and regional networks to the core and backbone networks. The business customers may need broadband network transport services to meet their demands for layer 1, layer 2, or layer 3 virtual private network (VPN) or application services. The transport technology is based on IP/MPLS over OTNs, and supports long haul transmission. However, standardized layer 2 Ethernet interfaces are used in all network domains to achieve cost efficient interconnections. Ethernet interface speeds of 1 Gb/s up to 10 Gb/s are currently required in the access and aggregation networks. The aggregated traffic transported in some trunk links of the core network already exceeds 40 Gb/s per wavelength channel, resulting in demands for higher transmission speeds.

IEEE Communications Magazine • March 2010

TOMIZAWA LAYOUT

2/17/10

Services

1:07 PM

Page 63

Access aggregation

Layer inter working

Layer 3 VPN services Layer 2 VPN services

Core backbone

Metro regional

IP/MPLS

IP/MPLS

Multi domain interworking

Layer 1 VPN services

OTN

Node

OTN

Figure 1. Transport network architecture.

Currently, the IEEE 802.3ba Task Force for 40GbE and 100GbE interfaces specifies higher speed Ethernet client interfaces for 40 Gb/s and 100 Gb/s rates, based on the following objectives, a standard is expected in June 2010: • Support full-duplex operation only • Preserve the 802.3 / Ethernet frame format utilizing the 802.3 MAC • Preserve minimum and maximum Frame Size of current 802.3 standard • Support a BER better than or equal to 10–12 at the MAC/PLS service interface • Provide appropriate support for OTN • Support a MAC data rate of 40 Gb/s • Provide Physical Layer specifications which support 40 Gb/s operation over: –at least 10km on SMF –at least 100m on OM3 MMF –at least 7m over a copper cable assembly –at least 1m over a backplane • Support a MAC data rate of 100 Gb/s • Provide Physical Layer specifications which support 100 Gb/s operation over: –at least 40km on SMF –at least 10km on SMF –at least 100m on OM3 MMF –at least 7m over a copper cable assembly IEEE 802.3ba physical media dependent (PMD) Ethernet interfaces are specified for the electrical backplane, copper, multi-mode fiber (MMF), and single-mode fiber (SMF) client interfaces, as shown in Table 1. For transport networks only SMF interfaces are of interest, since transmission lengths of several kilometers have to be supported. IEEE 802.3ba specifies the SMF interfaces using the multi-lane approach. The 40 Gb/s SMF interface is connected to a fiber pair using a 20 nm spacing CWDM grid in the 1300 nm wavelength range with 4 lanes of 10.3125 Gb/s per lane. The 100 Gb/s SMF interface is connected to a fiber pair using a 800 GHz spacing WDM grid in the 1300 nm wavelength range with 4 lanes of 25.78125 Gb/s per lane. In November 2009, an IEEE 802.3 study group on a serial 40 Gb/s SMF PMD Ethernet interface agreed to prepare 5 criteria, broad market potential, distinct identity, backward compatibility, technical feasibility, and economic

IEEE Communications Magazine • March 2010

PMD type

Lanes

Reach

Medium

40GBASE-KR4

4

1m

Electrical backplane

40GBASE-CR4

4

7m

Cu

40GBASE-SR4

4

100 m

MMF

40GBASE-LR4

4

10 km

SMF, ~1300 nm, 20 nm CWDM grid

100GBASE-CR10

10

7m

Cu

100GBASE-SR10

10

100 m

MMF

100GBASE-LR4

4

10 km

SMF, ~1300 nm, 800 GHz grid

100GBASE-ER4

4

40 km

SMF, ~1300 nm, 800 GHz grid

Table 1. IEEE 802.3ba 40 Gb/s and 100 Gb/s physical medium dependent (PMD) Ethernet interface types. K: Backplane, C: Copper, R: 64B/66B, S: Short, L: Long, E: Extended. feasibility, as the prerequisites to forming a task force (as 802.3bg). In contrast, OTNs require transmission over distances greater than 40 km for metro, regional, and backbone networks, as well as multi-channel DWDM transmission for higher link capacities in order to support scalable networks towards Tb/s capacities. The ITU-T Study Group 15 (SG15) standardized OTN for 40 Gb/s and 100 Gb/s transmission in Recommendation G.709, by specifying the optical channel transport units OTU3 and OTU4, respectively. These standardization bodies are closely linked by liaison activities and through the IEEE 802.3ba objective “provide appropriate support for OTN.” This is really essential, since the standards have to match to each other by supporting the same higher speed multi-lane structures for both types of 802.3ba and G.709 interfaces. This enables cost efficient Ethernet over OTN interfaces, since the structures and components can be reused. A good example is that both technologies use the same pluggable modules for several km single-mode interfaces; the bit-rates are slightly different due to the additional OTN overhead.

S29

TOMIZAWA LAYOUT

2/17/10

1:07 PM

Page 64

Ethernet is based on packet multiplexing, a LAN interface for less than 40km reach (stadardized as IEEE802.3).

Ether-network SDH is based on TDM, a WAN IF for legacy telecom network (standardized as ITU-T G.707. SDH-network Client

Multi client-signal accommodation

Server OTN is based on WDM, a WAN IF for photonic transport network (standardized as ITU-T G.709).

Optical transport network

1) Overhead Client

ODU

OTU format (16kB fixed-length frame)

3) FEC OTU

1

λ

3824 3825

16 17

4080

1 Fiber

Row 2 3

ODU: Optical-channel data unit OTU: Optical-channle transport unit

Column

Over head

4 1) Carrier-class OAM

Payload

2) Transparent mapping

FEC

3) High quality transmission

Figure 2. ITU-T G.709 optical transport network (OTN) functions.

TRANSPORT NETWORK’S REQUIREMENT AND ENABLING TECHNOLOGIES — EXTENDED OTN HIERARCHY AND DIGITAL COHERENT TECHNIQUE The introduction of higher speed 100 Gb/s transmission will not only increase the bandwidth of the interfaces and the throughput of the links, but it will also significantly impact the entire network. A seamless and less complex inter-working via standardized Ethernet interfaces has to be realized between the different network areas, which will increase the cost efficiency of the network by reducing the capital and operational expenditures. 100 Gb/s Ethernet over OTN will be used in the carrier network aggregation and transport areas, providing high quality broadband links with high granularity for the collected data flows; it supports different service classes and offers good network utilization. The cost efficiency of the transport networks have to be further increased by reducing the number of interfaces and transmission links between the network nodes. OTN is targeted as a unified infrastructure for multiple services as shown in Fig. 2, not only legacy telephone or SDH/SONET, but also Ethernet, Internet Protocol (IP), Fibre Channel and so on. The former standard of the carrier’s infrastructure, SDH/SONET, could transport Ethernet as well by the virtual concatenation (VCAT)

S30

technique, however its origin was still based on the telephone service of 64 kb/s. The main feature of SDH/SONET, its pointer processing, was set to decrease the amount of frame-buffer memories needed, which were expensive at that time. Memory cost has recently fallen dramatically due to the proliferation of computers and the Internet, and also the recent progress of semiconductor technologies. In that sense, OTN is more data oriented, since the transport mechanism uses many read/write actions from/to frame-buffer memories; the unit of read/write is still BYTE (8-bits) based. In the initial standard of G.709 approved in 2001 [1], the OTUk bit-rate was included mainly to satisfy the SDH/SONET requirement, a key client. Broadband services are now rapidly and widely spreading and Ethernet has taken a major role in not only LANs, but also in inter-office applications due to advances in optical interfaces. In the near future, high speed Ethernet will become the dominant client in terms of volume rather than SDH/SONET. In the longer term, it could be anticipated that SDH/SONET will gradually vanish since high-speed Ethernet can provide aggregation, switching, and interoffice transport. In this regard, it was proposed, by European and Japanese carriers (leadership by Deutsche Telekom and NTT), that OTN should be extended to better support Ethernet: the activity is called extended OTN. DT and NTT are proposing the extended hierarchy and the proposal is depicted in Fig. 3. The proposal is considered to has been triggered in 2006 time-

IEEE Communications Magazine • March 2010

TOMIZAWA LAYOUT

2/17/10

1x OTU4 111.81 Gb/s

(G.sup43) OTU3e

43.02 Gb/s (G.sup43)

OTU2e 11.09 Gb/s OTU2

Page 65

ODU4(L)

100GE

ITU-T SG15 is in

ODU4(H)

103.12 Gb/s

charge of the

2x

1x

standardization of

(G.sup43)

1x

44.57 Gb/s

OTU3

1:07 PM

ODU3e(H)

41.25 Gb/s

ODU3(L) ODU3(H)

1x

4x 4x

ODU2e(L)

STM256

frame-format, as well as some optical

10GE 10.31 Gb/s

ODU2(L)

1x

ODU2(H) 16x 4x 40x to ODU4(H)

1x

80x to ODU4(H) 32x to ODU3(H) 8x to ODU2(H)

ODU0(L)

ODU1(H) 2x

interfaces of OTN. It is actively discussing

STM64

and rapidly advanc-

9.95 Gb/s

ing the mapping of 1GbE/10GbE/40GbE/

STM16

ODU1(L)

the logical (electrical)

39.81 Gb/s

10x

10.71 Gb/s

OTU1 2.67 Gb/s

40GE

Transcoding

2.48 Gb/s

100GbE into the

1GE

OTN digital frame

1.25 Gb/s

format.

Figure 3. DT and NTT's proposal on the extended OTN hierarchy. (L) stands for lower-order ODU; the unit of client mapping and multiplexing, and (H) stands for higher-order ODU; the unit of transport. A detailed description is given in the article. frame, and the community will be generally in line with this approach, hence the hierarchy is likely to be approved in December 2009. On the other hand, from the viewpoint of optical transport technology, carriers’ requirements lead to a completely different technological paradigm. Carriers have already commenced the deployment of 10 Gb DWDM links, with 40 Gb transponders sometimes being deployed in 10 Gb designed links. Hence a strict design requirement in 40 Gb system development was compatibility with 10 Gb designed links; i.e., 50 GHz/100 GHz spacing and Reconfigurable Optical Add Drop Multiplexer (ROADM) cascade, the same span of in-line amplifiers, the similar level of Differential Group Delay (DGD), and so on. It is noted that the most important requirement for transport networks is cost-efficient operation, which is closely connected to network scalability. Only future proof scalable network technologies like Ethernet over OTN DWDM networks, scaling up to e.g., 8 Tb/s with 80 × 100 Gb/s DWDM channels on a 50 GHz grid resulting in a spectral efficiency of 2 bit/s/Hz, can manage the expected 10 fold increase in traffic in 5 years. Since the same requirements must be satisfied by 100 Gb transport systems, we need extremely innovative solutions. One such solution is digital coherent detection; a polarization-multiplexed multilevel encoded signal is detected by a coherent receiver with local oscillator (LO) with 90 degree hybrid interferometer. The OE converted signal is subjected to Analogue-Digital Conversion (ADC) and multi-bit signals are launched into a Digital Signal Processor (DSP) which compensates the waveform distortion created by Chromatic Dispersion (CD) and Polarization Mode Dispersion (PMD), other inter-symbol interference, and frequency mismatch between transmitted signal and LO. Since the baud-rate is decreased by multilevel coding and polarization multiplexing, into 25 Gb (up to around 30 Gb due to strong FEC redundancy) or by moreover two sub-carriers

IEEE Communications Magazine • March 2010

solution to 12.5 Gb (15 Gb), the spectral width well matches the requirement, and in addition, dispersion (CD and PMD) tolerance is drastically improved by the sophisticated electric equalization provided by the DSP. A technical assessment of the 100 Gb digital coherent systems enabled by a set of new technologies is described later. Other requirements for Ethernet over OTN architectures include a pan-layer network management and operation system, administration and maintenance (OAM) signal flows between the OTN, Ethernet, and IP layers to support transport network optimization, resiliency and protection switching e.g., for failure detection, forwarding, and localization.

EXTENDED OTN AND ITS STANDARDIZATION IN ITU-T SG15 ITU-T SG15 is in charge of the standardization of the logical (electrical) frame-format, as well as some optical interfaces of OTN. It is actively discussing and rapidly advancing the mapping of 1GbE/10GbE/40GbE/100GbE into the OTN digital frame format. The digital frame consists of 4-bytes row × 4080-bytes column, and this structure is identical for all bit-rates (the frame period shortens as the bit-rate rises). The frame, called Optical channel Transport Unit-k (OTUk), includes Optical channel Data Unit-k (ODUk), and Optical channel Payload Unit-k (OPUk); k represents the level in the hierarchy: OTU1 = 2.67 Gb/s, OTU2 = 10.71 Gb/s, OTU3 = 43.02 Gb/s, and OTU4 = 111.81 Gb/s. From 2004, extensive discussions were conducted on 10GbE-LAN (10.3125 Gb/s) mapping into ODU2/OTU2. Initially, IEEE specified 10GbE-WAN (9.95328 Gb/s) to comply with SDH/SONET STM-64/OC-192 signal. However, the market pushed 10GbE-LAN as a cost-effective interface resulting in the dominance of 10GbE-LAN as a router/switch interface. Carriers should now transport 10GbE-LAN signals

S31

TOMIZAWA LAYOUT

2/17/10

The background of this proposal is the cost effectiveness of the components created in response to the single agreement on modulation format. 100 Gb DP-QPSK is currently proposed and a discussion is underway among several system vendors and component vendors.

S32

1:07 PM

Page 66

over their infrastructure (OTN). Several different requirements on the transparency of client 10GbE-LAN signal have appeared, contradicting each other. If bit-level transparency is required, the resultant bit-rate of OTU2 has to be overclocked. However, if the bit-rate must match the existing standard, some portions of the client signal should be terminated (preamble, SFD, or IFG are sometimes used in commercial switches/routers). The latter is called information (MAC)-transparency. ITU-T issued Supplement document (G.Sup43) that describes a list of implementations on 10GbE-LAN over ODU2/OTU2 and ODU2e/OTU2e (e stands for overclocked bit-rate) [2]. It is noted that using overclocked 44 Gb to carry 4 × 10GbE-LAN (indicated as ODU3e/OTU3e) has been also agreed and is described in G.Sup43. It should be noted, in ODU3e/OTU3e, that the client signal is especially focused on 4 × 10GbE-LAN signal, and the mixture of the other clients (e.g., STM64, etc.) is not agreed in G.Sup43 although it appears to be technically possible. There are two ODU3e/OTU3e types, namely ODU3e1 based on Asynchronous Mapping Procedure (AMP) and ODU3e2 based on Generic Mapping Procedure (GMP), introduced later in this section. Each of them has a mechanism to offset the frequency deviation ±100 ppm of 10GE-LAN (via ODU2e) into ±20 ppm OTU3e signal during time-division multiplexing, this mechanism is called frequency justification. ITU-T takes the position that the supplement is not actually a standard, merely information. For 40GE, while IEEE defines 64B/66B based bit-rate for 40GE (4 × 10.3125 Gb/s) again, the transcoding approach is proposed for maintaining the existing bit-rate of ODU3/OTU3, where 64B/66B coding is decoded and re-encoded by 512B/513B based coding (more precisely 1024B/1027B with 3-bit headers to improve Mean Time To False Packet Acceptance [MTTFPA]). The transparency level is called code-word transparency, a compromise between bit transparency and information (MAC)-transparency. For 100 Gb OTN transport, the ODU4/OTU4 hierarchy has been agreed with the bit-rate of 111.809974 Gb/s (= 255/227 × 2.488320 Gb/s × 40). The bit rate was chosen in order to provide 80 tributary slots (TS) for up to 80 ODU0 and to support up to 10 ODU2e. A division of 3808 columns by 80 TS results in an odd number of 47.6 columns per TS. Columns 3817 to 3824 are, therefore, fixed stuff. The remaining 3800 columns divided by 80 results in 47.5 columns per TS. The minimum bit rate of a TS is determined by the ODU2e bit rate. ODU4s should run with an STM-16 based clock of 2.488320 Gb/s, as other ODUs do (ODU2e is an exception). The best fit was the quotient 238/227 for the OPU4, i.e., 255/277 for the OTU4. Multiplexing of 10GbE-LAN has been agreed as bittransparent mapping/multiplexing using ODU2e(L) into ODU4(H) for the first time as a standard. Although this extension has a big impact on the existing OTN infrastructure, and indeed there were several opposing proposals, ITU-T is likely to approve the extended OTN hierarchy in December 2009, shortly before the

approval of IEEE 40GbE/100GbE, thanks to the strenuous efforts of the delegates from many organizations. Not only the hierarchy, but also the mapping procedure is extended in the new G.709 for including GMP. Formerly, Bit-synchronous Mapping Procedure (BMP) and Asynchronous Mapping Procedure (AMP) were defined in the original standard. In BMP, the line signal clock is synchronized to the incoming client clock. Hence, if the client signal (for instance 10GbE) has a frequency deviation of ±100 ppm, the line clock has the same amount of deviation ±100 ppm. In AMP, especially used in the ODU multiplexing regime, the frequency deviations of multiple clients are justified into a precise line clock by using both positive and negative stuffing bytes. As proposed, GMP is characterized by positive stuffing. The number of client bytes (and bits) mapped in the payload of ODUk/OTUk is transferred by using overhead bytes; this information is used to regenerate the client clock at the line receiver side. GMP is more tolerant than AMP to the frequency deviations of the client clocks, i.e., GMP is more elastic to bit-rate differences in the client signals. ITU-T SG15 is also in charge of standardizing the optical physical interface in OTN, especially for Inter-Domain Interfaces (IrDI) which are defined as carrier-to-carrier connections as well as multiple vendor compatible interfaces (called as transverse compatible interfaces). As the DWDM (OTN) line section consists of single vendor products, for long-haul DWDM (OTN) optical interfaces, more detailed specifications are out of scope of ITU-T SG15, and are up to the vendor or some form of joint engineering. ITU-T SG15 will define only a small set of requirements of optical interfaces (frequency grid, channel spacing, maximum dispersion, and repeater spacing, etc.) [3].

TECHNICAL ASSESSMENT OF DIGITAL COHERENT SYSTEMS FOR REAL 100GBE/OTN TRANSPORT The Optical Interworking Forum (OIF) specifies higher speed transport issues in the “100 Gb long haul DWDM” project. The framework of the multi-source agreement describes functionalities for dual polarization — quadrature phase shift keying — coherent detection (DPQPSK-CD) modulation formats, forward error correction (FEC), integrated photonics, and electro mechanicals for transceiver modules that are cost efficient and pluggable [4–8]. It is proposed to take the single solution of the 100 Gb DWDM modulation format, even in IntraDomain Interface (IaDI). The background of this proposal is the cost effectiveness of the components created in response to the single agreement on modulation format. 100 Gb DPQPSK is currently proposed and a discussion is underway among several system vendors and component vendors. For FEC, ITU agreed FEC on IrDI (across multiple carrier/vendor islands) as ReedSolomon (255, 239) as is defined in G.709 for lower speeds, and FEC of IaDI (within a single carrier/vendor island, such as DWDM links) is

IEEE Communications Magazine • March 2010

2/17/10

1:07 PM

Page 67

likely to be up to the vendor or some form of joint engineering. Since 100 Gb DWDM requires 4 dB more margin than 40 Gb DWDM, care is needed to find this gain, possibly in the receiving technology and/or FEC. It should be noted that, however, the performance of 40 Gb Enhanced FEC (EFEC) is already close to the theoretical limit under 7 percent redundancy and hard decision regime. We may have to get a new paradigm. Since the digital coherent receiver includes Analog-to-Digital Conversion (ADC) followed by a Digital Signal Processor (DSP), in the ASIC, there are already a multiple number of bits representing an instantaneous sample of the incoming signal, which is directly useful to Soft-Decision FEC. In that sense, it may be beneficial to integrate SD-FEC into the DSP-LSI. There is an extensive discussion in OIF about the use of EFEC (or SD-FEC) in the long-haul OTN line-side. However, no agreement has been obtained on the standardization of FEC code since FEC selection and design are considered to be left to the vendor to drive competition. So far, the standard redundancy of 7 percent and extra redundancy of about 20 percent have been proposed. Within the digital coherent transport systems, the most challenging development issue is, of course, realizing the electric ADC and DSP. Assuming the DP-QPSK format, the ADC-DSP interconnection is 1.3 Tb/s throughput (28 Gbaud × 4 lanes × 2 samples × 6 bits (for ADC)). The multiple chip approach seems unlikely and an integrated single chip is the challenging but expected solution; it will be enabled only by advances in CMOS technology. CMOS semiconductor process development is usually driven by big market needs; i.e., mobile phones, game instruments, and so on, and not by relatively small markets like optical transport. Following the CMOS technology trend for ADC, according to International Technology Roadmap for Semiconductors (ITRS) [9], the RF and analog mixed signal processing speeds of CMOS devices is increasing by the yearly factor of 1.15–1.2 as is depicted in Fig. 4. Provided the volume production of 43 Gb based ADC+DSP devices became mature in 2006, the same level of production seems possible for ADC+DSP devices for 112 Gb transport by 2012. A similar result can be obtained from Walden’s law for ADC [10], indicating that the next generation will appear in 2013. That roadmap, coincidentally, matches the market forecast of 100GbE volume ramp-up. It is noted that the above estimation will be applicable to regular and mature volume production, while earlier development may be feasible by different approaches to meet the requirements of earlyadoption carriers. For early adopters, interim approaches could be used until the achievement of standard and very high-speed ADCs in CMOS and their integration in very large DSP circuits. One approach would be the multiple chip solution with hybrid integration yielding single packages. Another approach is advance special sampling technologies applicable to mature and sophisticated CMOS process. Early 100 Gb DP-QPSK digital coherent transport systems may be introduced in the latter half of

IEEE Communications Magazine • March 2010

7

Peak Ft (GHz) Peak Fmax (GHz)

RF and analog CMOS

6 Ratio with respect to 2007

TOMIZAWA LAYOUT

5

4

3

2

1 0 2007 ‘08

‘09 ‘10

‘11 ‘12

‘13

‘14

‘15

‘16

‘17

‘18

‘19 ‘20

‘21

Year

Figure 4. Technology roadmap in RF and analog signal processing speed in CMOS (according to ITRS). Ft: current gain transit frequency; Fmax: maximum oscillation frequency. the 2010 to 2011 time frame (note that other solutions may exist using in-coherent or multiple sub-carrier technique, and their deployments may be earlier than single carrier DP-QPSK coherent system).

FUTURE (400GBE/OTN, OR 1TBE/OTN) Since 40GbE/100GbE and related extended OTN will be completed in 2010, discussions on the future bit-rates must be facilitated worldwide. Requirements from the data-center seem special, where 1Tb/s seems preferable for such kind of applications as there simply isn’t enough fiber to even go from one floor to another floor. On the other hand, for the transport side, technical barriers to overcome are so critical that intermediate step seems necessary; 400 Gb/s may be the next candidate. Care should be taken as to the timing of the decision because it depends on the state-of-the-art technologies, and also because the market maturity for 40 Gb/100 Gb should be well established before diving into the next paradigm to allow carriers to enjoy the lowcost and high-performance systems. A simple and historical extrapolation shows that 400GbE/1TbE may appear in or around 2017, considering 10GbE and 40GbE/100GbE appearances in 2003 and 2010 respectively.

CONCLUSION This article overviewed requirements, standardization activities, and enabling technologies regarding the evolution of the Optical Transport Network (OTN) with 100 Gigabit Ethernet. Joint requirements from the R&D activities of two international carriers were described. Since high-speed Ethernet is taking a more important role in the LAN, access, and aggregation than in the past, the infrastructure OTN should be

S33

TOMIZAWA LAYOUT

2/17/10

Since high-speed Ethernet is taking a more important role

1:07 PM

Page 68

extended to better support Ethernet. Carriers’ requirements for scaling networks will lead to innovative digital coherent transport systems, and a technical assessment of a realistic roadmap was also described.

in the LAN, access,

ACKNOWLEDGMENTS

and aggregation

The authors thank Arnold Mattheus for his efforts, as the contact person in this carrier-tocarrier collaboration. The authors thank Takuya Ohara, Shigeki Aisawa, and Mitsuhiro Teshima for their support, Tetsuo Takahashi, Yutaka Miyamoto, Shinji Matsuoka, and Takashi Nakashima in NTT Network Innovation Labs, and Kazuo Hagimoto, director of NTT Science & Core Technology laboratory group, for their continuous encouragement. Furthermore, the authors would like to thank Bruno Orth and Hans-Dieter Haag from the Deutsche Telekom Headquarter, and Andreas Gladisch from the Deutsche Telekom Laboratories for their continuous support of the work. This work is supported in part by the National Institute of Information and Communications Technology (NICT) of Japan under “Lambda-Access Project” and “Universal Link Project”. A part of this work belongs to “Next-generation High-efficiency Network Device Project” which Photonics Electronics Technology Research Association (PETRA) contracted with New Energy and Industrial Technology Development Organization (NEDO).

than in the past, the infrastructure OTN should be extended to better support Ethernet.

REFERENCES [1] ITU-T Rec. G.709/Y.1331, “Interfaces for the Optical Transport Network (OTN).” [2] ITU-T Supplement G.Sup43, “Transport of IEEE 10G base-R in Optical Transport Networks (OTN),” Nov. 2006. [3] ITU-T Rec. G.696.1, “Longitudinally Compatible Intradomain DWDM Applications.” [4] OIF, “100G Ultra Long Haul DWDM Framework Document.” [5] OIF, “100G FEC White Paper,” to be published. [6] OIF, “Implementation Agreement for Integrated Polarization Multiplexed Quadrature Modulated Transmitters,” to be published. [7] OIF, “Implementation Agreement for Integrated Dual Polarization Intradyne Coherent Receivers,” to be published. [8] OIF, “Implementation Agreement for 100G Long-Haul DWDM Transmission Module,” to be published. [9] International Technology Roadmap for Semiconductors; http://www.itrs.net/ [10] R. Walden, “Analog-to-Digital Converter survey and analysis,” IEEE JSAC, vol. 17, no. 4, Apr. 1999, pp. 539–50.

S34

BIOGRAPHIES JOSEF ROESE is Senior Project Manager at Deutsche Telekom AG, Headquarter. He received his Dipl. Ing. at the University of Applied Science in Muenster, Germany in 1981 and joint Deutsche Telekom in 1983. In the past 18 years, he works for a number of organization units of Deutsche Telekom on the field of transport networks. The activities covered evaluation of transport network technologies and network architectures. He participated in the standardization of SDH and since the end of the 1990s in the standardization of OTN. As such, he contributed to Recommendation G.709 versions 1, 2 and 3. He his also engaged in IEEE 802.1. R ALF -P ETER B RAUN is Senior Project Manager at Deutsche Telekom AG, Laboratories, leading R&D projects for next generation networks. He received his M.S. and Ph.D. in Electrical Engineering from the Technical University Berlin, Germany, in 1985 and 1995, respectively. He has more than 20 years of research and industry experience on optical transmission systems and network architectures. After 14 years of advanced research work at the Heinrich-Hertz-Institut, Berlin, Germany, he joined the Deutsche Telekom AG in 1997. He is engaged in the topics of network architecture, technology, service, control, and standardization. He is a member of VDE, ITG, IEEE802.3 working group, MEF, and PMI. M A S A H I T O T O M I Z A W A [M‘92] (tomizawa.masahito@ lab.ntt.co.jp) is a senior research engineer, supervisor, at Nippon Telegraph and Telephone Corporation (NTT). He received M.S. and Ph.D. in Applied Physics from Waseda University, Tokyo, in 1992 and 2000, respectively. From 2003 to 2004, he was on leave to Massachusetts Institute of Technology (MIT), Cambridge, MA, USA. In 1992, he joined NTT R&D and since then, he has been engaged in R&D for high-speed optical transmission systems and their deployments, as well as international standardization activities in ITU-T Study Group 15, and also international carrier-to-carrier collaboration for several years. In 2006 he received the young scientist award from The Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan. In 2007 he received Sakurai-Kenjiro memorial award from Optoelectronic Industry and Technology Development Association (OITDA) of Japan. O SAMU I SHIDA [M‘88] ([email protected]) is a senior research engineer, supervisor, at Nippon Telegraph and Telephone Corporation (NTT). He currently leads photonic networking systems research group in NTT Network Innovation Labs, Yokosuka, Japan, and is responsible for the research on architecture and interfaces of converged packet/optical transport network. He has over 20 years of experience at NTT Laboratories in research on media networking systems, high-speed Ethernet transport, optical cross-connect systems, WDM systems, coherent optical fiber communications and their subsystems employing planar lightwave circuits (PLC) and tunable diode lasers. Also he was involved in the development of several Ethernet standards (IEEE802.3ae, IEEE802.3ba) as well as in the revision of the OTN standard (ITU-T Recommendation G.709). He holds a B.E. and M.E. in electrical engineering from the University of Tokyo, Japan. He is the author or coauthor of more than 50 journal and conference articles (in English), and the co-editor of a textbook (in Japanese) on 10 Gb/s Ethernet Technologies (Impress, 2005).

IEEE Communications Magazine • March 2010