Sponsored By

Long-Haul WDM: RFI Exercise

Four vendors put forward proposals for a pan-European network * 40 Gbit/s vs 10 Gbit/s * Optical vs Electrical * Risk vs The Safe Play

March 12, 2002

19 Min Read
Long-Haul WDM: RFI Exercise

When vendors bid for a big contract to build a long-haul network, their proposals can provide useful insights into the current state of development of transmission and switching technologies.

Usually, of course, only the carrier concerned gets to see what's on offer. But last year Light Reading staged an exercise in which four vendors responded to a simplified RFI (Request for Information) for part of a hypothetical pan-European backbone dubbed Euro 1. Their proposals were presented at our Lightspeed Europe conference in December.

Stuart Barnes, entrepreneur in residence at Atlas Venture and a former guru of transmission technology at Alcatel SA (NYSE: ALA; Paris: CGEP:PA), put together the original RFI and chaired the conference session. Afterwards, he provided an analysis of the vendors' proposals, which is presented in this report.

First off, Light Reading would like to thank Stuart Barnes for all the work he put into this project and thank the vendors for being good sports and participating in this exercise. The individuals involved were:

Here's a hyperlinked summary of the report:

— Introduction by Craig Williamson, Associate Editor, Light Reading
http://www.lightreading.comIn an atypical act of subtlety, Light Reading set a challenge that was designed to explore the technical direction of the four participating organizations over the next three to four years. With reduced cost-per-unit bandwidth being the driver, the industry is currently on the cusp of two major new technologies – namely, 40-Gbit/s Wavelength Division Multiplexing (WDM) transmission and network transparency. The evolution towards these solutions is charted below (graphics courtesy of Nortel). This evolutionary trail illustrates how solutions moved from 2.5 Gbit/s to 10 Gbit/s on each Optical Fiber, before the introduction of WDM technologies that allowed many separate 10-Gbit/s signals to be transmitted on just one fiber. More recently, 10-Gbit/s WDM signals have been able to travel over several thousands of kilometers due to the establishment of Optical Amplification and the advent of optical add/drop multiplexers allowing a certain degree of transparency.In 2002 we are at another significant marker in the industry, with 40 Gbit/s emerging as a possible technology for deployment in today’s networks, perhaps alongside or as a replacement for existing 10-Gbit/s WDM. The cost trend illustrated below indicates how the industry is continually squeezing as much value for money as possible out of transmission equipment, making a move over to these new technologies an increasingly attractive proposition. The Euro 1 challenge was designed to explore the respondents' views on the readiness of operators to cross these divides. The results were both surprising and illuminating.The RFI

The fictitious Euro 1 is part of a larger pan-European Network based on the interlinking of a set of physical rings. The ring in question is typical of those that have sprung up in recent years, with a mixture of constructed and leased fiber. The fibers are a mixture of standard singlemode fiber (SMF) and large effective-area fiber (LEAF) (see Advanced Fiber Types for further details). The lengths of the connections were deliberately chosen to challenge the bandwidth and distance capabilities of the vendors. This is particularly relevant in Europe, where regenerative spans have been modest (to put it mildly) to date. In order to accentuate the possibilities, we also concocted a highly ambitious growth profile for the links, shown on the diagram by 10-Gbit/s channel requirements at the ends of years 1, 2, and 3.To accommodate this growth, the vendors were forced to decide the most cost-effective way forward: either more fiber; better spectral efficiency through higher time-division multiplexing (TDM) rates; or closer channel spacing.Finally, we created two cross-connection sites (nodes B and E) with significant traffic flows, in order to see which vendors were considering full transparency at these sites in order to minimize costly regeneration.Other relevant data provided for the exercise:

  • Year 1 – 50:50 Voice:Data, Year 3 – 10:90 Voice:Data

  • Links AB and BC

    • Operator-owned SMF

    • Loss = 0.25 dB/km

    • Polarization mode dispersion (PMD) = 0.25 ps/sq rt km

    • 20dB span loss between amplifier points

  • All other links

    • Operator-leased LEAF

    • Loss = 0.25 dB/km

    • PMD = 0.05 ps/sq rt km (picoseconds per square-root kilometer)

    • 22dB span loss between amplifier points

The following four pages set out the responses from suppliers:

Presented by Ian Clarke, Senior Systems Engineering Manager of EMEA, CienaThe solution presented by Ciena Corp. (Nasdaq: CIEN) is built around a mesh topology with a number of 10-Gbit/s wavelengths that increases as the system evolves, with tight spacing of the wavelengths removing the need for additional fibers.Ciena’s CoreStream DWDM terminals can send and receive the 10-Gbit/s streams via wavelengths with 50GHz (0.4 nanometer) spacing. These wavelengths can be in either the C-band, (approximately 1530 nm to 1580 nm), the L-band, (around 1580 nm to 1610 nm), or a combination of each. The Year 1 solution would use Tunable Lasers with 50GHz (0.4 nm) spacing, possibly moving towards 12.5 GHz (0.1 nm) as the technology matures and the demand increases for Year 3.The CoreStream amplifiers in the system are gain-flattened Erbium Doped-Fiber Amplifiers (EDFAs) that can provide even amplification across the range of wavelengths being transmitted. They are combined with Dispersion-compensating functions and have the option of adding Raman Amplification for further reach.The switching workhorse is Ciena's electrical CoreDirector crossconnect. This has the ability to groom lower rate traffic while supporting up to 64x10-Gbit/s interfaces in its initial single bay configuration and being ready to adapt for 40-Gbit/s signals in the future. It can also provide network protection in mesh, ring, or linear configurations. Secondary CoreDirectors are needed in Year 3, due to port density requirements.Further details:

Presented by Peter Ball, VP, Network Solutions, FujitsuFujitsu Networks Europe Ltd. and Tellium Inc. (Nasdaq: TELM) have proposed a joint solution to the Euro 1 requirements. The solution is based around a mesh system with 10-Gbit/s technology and, without plans for more densely packed wavelengths, this inevitably requires more fibers to be installed.The Fujitsu Flashwave OADX provides network transport. The system can be scaled up to 176x10-Gbit/s channels while in service and supports both the C-band and L-band. At its maximum capacity of 1.76 Tbit/s, transmission over 600 km is possible. Reducing this to 0.8 Tbit/s enables distances up to 1,500 km. Gain-flattened Erbium Doped-Fiber Amplifiers (EDFAs) are used, and there is the option to add Raman Amplification for longer links.Tellium’s Aurora Optical Switch uses an electrical switching matrix, although true optical switching is planned for the future (see Optical Crossconnects). It can handle up to 128x10-Gbit/s ports in units that can be combined to support thousands of ports with total capacities exceeding 20 Tbit/s. Grooming at the 2.5-Gbit/s level is also possible.It was assumed that protection was required for all traffic, and therefore the crossconnect equipment implements 1+1 protection for voice traffic and mesh restoration for data traffic. The optical add/drop multiplexer implements 1+1 protection for all traffic. Details of the cost and usable capacity benefits of mesh protection were given, relating to the Fujitsu/Tellium proposal as it evolves through years 1, 2, and 3.Further details:

Presented by Simon Wolting, Senior Network Architect, LucentThe Lucent Technologies Inc. (NYSE: LU) solution begins with 10-Gbit/s technology in Year 1, but soon introduces an additional 2.56 Tbit/s of capacity in Year 2, courtesy of 40-Gbit/s wavelengths. These higher-capacity wavelengths are used to provide an express route between the high-capacity crossconnect nodes B and E, as well as providing additional capacity to node A.The transmission solution is Lucent’s LambdaXtreme Transport system. This can perform remotely configurable optical add/drop multiplexer (OADM) functions, while also allowing express traffic to pass through without the need for electrical Regeneration. Both 10-Gbit/s and 40-Gbit/s wavelengths can be handled on the same platform, and Raman Amplification is used to bridge ultra long haul distances up to thousands of kilometers without regeneration.At each crossconnect node sits Lucent’s LambdaRouter All Optical Switch. This is a true optical switch based upon micro-electro-mechanical systems (MEMS) (see Optical Crossconnects). True optical switching means that any bit-rate or protocol of signal can be switched between fibers, as the switch fabric itself is simply rerouting light and not actually interpreting it as information. The system also provides overall mesh protection for the Euro 1 network and can quickly provision new wavelengths as required.Further details:

Presented by Serge Melle, VP EMEA, Market Development, Optical Internet, Nortel NetworksThe Nortel Networks Corp. (NYSE/Toronto: NT) solution begins with 10-Gbit/s wavelengths but moves to 40 Gbit/s in years 2 and 3. The higher-capacity wavelengths are primarily used as an express route between the two high-capacity cross-connection nodes, B and E. Data can be transmitted through these express routes without the need for electrical regeneration in between, although Raman Amplification may be required.Nortel’s OPTera LH (long haul) series of optical line systems transmit the data and can provide amplification through Erbium Doped-Fiber Amplifiers (EDFAs)as well as Raman amplification. They can support up to 160 wavelengths over a single fiber, and there is even an upgrade path to 80-Gbit/s wavelengths in the future.The OPTera Connect DX optical switch provides electrical add/drop and traffic grooming functionality, scaleable up to 320 Gbit/s. The HDX gives large-scale electrical crossconnect capabilities at the main nodes. It can handle hundreds of wavelengths of 2.5-, 10-, or 40-Gbit/s each, and can scale to a total capacity of 40 Tbit/s.The future for these main nodes would be the OPTera Connect PX Photonic Switch that provides all-optical switching capability based upon micro-electro-mechanical systems (MEMS) (see Optical Crossconnects). This allows wavelength switching for through traffic without the need for electrical processing and can also work in conjunction with the electrical switches to perform grooming below the wavelength level as well other electrical domain functions.Further details:

We've analyzed the results in two parts. First of all, we shall look at the transmission features of the vendors’ responses as summarized in the table below.
Table 1: Transmission Features

Ciena

Fujitsu/Tellium

Lucent

Nortel

Route to high bandwidth

Closely spaced
Nx10 Gbit/s

More fibers
Nx10 Gbit/s

Nx40 Gbit/s

Nx40 Gbit/s

Distance (terrestrial)

< 1000 km

800 km, 1.7 Tbit/s
1,500 km, 0.8 Tbit/s

1000 km release 1
3000 km release 2

1200 km

Channel spacing

50 GHz now
12.5 GHz in 3 years

100 GHz
moving to 50 GHz

100 GHz

200 GHz

Amplifier spacing

~ 100 km for ultra long haul

100 km

100 km

Details not presented

Amplifier type

C Band, L Band, Raman

C Band

L Band, Raman

C Band,
L Band, Raman

Dispersion management

At electrical add/drop multiplexers and crossconnect points

At electrical add/drop multiplexers and crossconnect points

Modules being tested in the lab

Details not presented



A cursory glance shows that two camps have been formed: Ciena and Fujitsu/Tellium are betting the mortgage on 10-Gbit/s Wavelength Division Multiplexing (WDM) technology being impregnable over the next two to three years, whilst Nortel and Lucent believe that 40-Gbit/s technology can be driven through soon, to realize a significant cost benefit to operators. Nortel produced historical evidence supporting these claims (see The Challenge). So, with what high stakes are these big rollers playing? In the case of Ciena and Fujitsu/Tellium, the strategy appears to be woven around their current preeminence in electrical crossconnects. They are the leaders of the pack at present, but of course the switch fabrics (despite market-speak) are fairly inflexible in terms of going from 10 to 40 Gbit/s. So why bet the house on 40 Gbit/s?The Fujitsu/Tellium play was by far the safest bet. The technology proposed was easily the shortest stretch and they may be right. With a surfeit of fiber out there, why not use it? The Ciena play was a little riskier but not much. Despite claims that it was an easy play to go to 12.5GHz channel spacing over 600 km on legacy links, we were not able to test this in the limited time we had before the organizers literally pulled the plug.Nortel and Lucent do not have such strength in electrical crossconnects. Furthermore, both have had access to huge component research capabilities and Nortel still does. It is therefore of little surprise that they would want to raise the stakes and not only disrupt the 10-Gbit/s apple cart but also the stranglehold of Ciena and Fujitsu/Tellium in crossconnects.But at what risk? To achieve 40 Gbit/s on the network scale described is not just a case of developing suitable line terminal equipment (and this is no trivial task by any standards) but with Nx40-Gbit/s, system engineering issues are in your face. In addition to the 40-Gbit/s components for long haul (and there were some teething problems for 10 Gbit/s), linear phenomena such as Polarization Mode Dispersion (PMD) and Nonlinear Effects such as four-wave mixing become critical. Chromatic Dispersion mapping will be essential to achieve the transparent distances claimed by Nortel and Lucent. These are meaty system engineering questions, so how will this be done – and with what components – to maintain a downward price trend?All the respondents were in favor of Raman Amplification, and it looks as if there are no further safety worries out there. Lucent gave an interesting twist by favouring the L-band for amplification in the future, more or less abandoning the C-band. The L-band is broader, and there are some merits in this approach, but the availability of colored sources for Nx40 Gbit/s is somewhat uncertain at present.And now onto the network elements proposed by each vendor. The table below summarizes the main features of the respondents’ add/drop and crossconnect technologies.
Table 2: Network Elements

Ciena

Fujitsu/Tellium

Lucent

Nortel

Network

  • Mesh

  • Nx10 Gbit/s

  • Grooming

  • Mesh

  • Shared protection feature for low system cost

  • Can support evolution to 40 Gbit/s

  • Mesh

  • 40-Gbit/s optical express for connecting rings

  • Optical crossconnect to minimize electrical regeneration

  • Mesh

  • 40-Gbit/s optical express for connecting rings

  • Optical crossconnect to minimize electrical regeneration

Add/Drop

CoreStream

Flashwave OADX

LambdaXtreme

OPTera DX

Electrical

Electrical

Optical for through traffic

Optical for through traffic

Electrical for drop

Electrical for drop

Strong grooming features

Designed for through traffic

Designed for through traffic

Modular and scaleable

Expandable to 1.76 Tbit/s

Common 10-Gbit/s and 40-Gbit/s platform

Low power consumption

Crossconnects

CoreDirector

Aurora Optical Switch

LambdaRouter

OPTera HDX and PX

Electrical

Electrical

Optical

Electrical (HDX) moving to Optical (PX)

Strong grooming features

Scaleable to 20 Tbit/s
Upgradeable in service

Common 10-Gbit/s and 40-Gbit/s platform

PX supports band switching

Modular and scaleable to > 7 Tbit/s

Self-tuning to enable turnup/ provisioning

Protocol independent

HDX supports grooming

No regeneration required in entire ring for express traffic

HDX supports 2.5-Gbit/s and 10-Gbit/s ports



Again we see the broad divide. Ciena and Fujitsu/Tellium stick to their knitting, whilst Nortel and Lucent embrace the newer concepts of optical transparency. The Ciena and Fujitsu/Tellium plays are safe once more, but at what cost to the operators? After the major expense of building the fiber infrastructure, Regeneration costs are the major transmission outlay for operators, and these vendors offer little good news on this front.The Nortel and Lucent plays offer the promise of a much more logical network for operators in future years, where photons are used for transmission and electrons for grooming features at the edge of the network. This is arguably the most efficient application for these elements, and such a design was illustrated in Lucent's presentation with the graphic below.But are the operators ready to take this stride forward, and are all the smart functions available to enable these networks? Since the breakdown of the PTTs, there has been minimal cohesion in the standards process: The last comprehensive standards process was the Sonet (Synchronous Optical NETwork) and SDH (Synchronous Digital Hierarchy) initiative, whereby both the physical and the management infrastructures were jointly addressed. There have been pockets of standards activities, most notably with the advances of Multiprotocol Label Switching (MPLS), but this is not nearly as comprehensive as required to introduce a complete, transparent network.Disruptive technologies since then have therefore relied on early adopters, such as those that rolled out Wavelength Division Multiplexing (WDM). MCI (Nasdaq: MCIT), Sprint Corp. (NYSE: FON) and Williams Communications Group (NYSE: WCG) all benefited from this approach. But this was a point-to-point system and did not depend on the resolution of some fairly serious management and monitoring issues. There is a significant difference between the testbed requirements for qualifying an optical network concept and those of a point-to-point WDM solution. Furthermore, there has been insufficient work carried out in the various international research programs (MONET, ACTS, CANARIE, etc.) to argue that these products can be rolled out into network solutions immediately and with the confidence that followed the Sonet/SDH standards and trailing activities. So how will this pan out? Are there any early adopters left? Do they have the test infrastructures to fully test and approve these products? What will the timescales be to get comfortable with the technology? I have been a strong advocate of true optical networks over recent years because of the disruptive potential of the concept. In my opinion, it is the only solution out there that will offer a disruptive price/performance benefit to the consumer. I do have my concerns that all the pieces are not yet in place. The standards process is not sufficiently comprehensive, nor do there appear to be any early adopters out there with the network test capabilities to approve these products in the short term. And do they have the appetite at the moment?I hope that Nortel and Lucent can prove me wrong, but we had little time at the conference to dig into these issues. My great concern, too, is that the worldwide debate on the control-plane issues has distracted operators and vendors alike from tackling many of the practical issues involved in constructing such networks. After all, it will no longer be digital!And finally, who is the weakest link? The ultimate arbiters on this will be the operators – when they finally stick their heads over the parapet. Do they follow the safe play and go for the electrical solution? Or do they try and build future-proof photonic networks based on transparent wavelength switching, thus reaping the longer-term financial benefits? Time will tell. What is dead certain is that there will be some big winners and big losers.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like