Technical conferences often provide a good indicator of what “the next big thing” in optical networking will be, and this year’s European Conference on Optical Communications (ECOC) in Brussels was no exception. One of the highlights of the event was from Verizon, which presented the results of a field trial with Nokia Siemens Networks showing that the vendor’s 100 Gbit/s solution could overlay existing 10 and 40 Gbit/s traffic on fibre that had been specially selected for its poor transmission performance. Shortly afterwards, the US operator announced results of a second, similar field trial, this time with Nortel. “Each advance we make moves the Verizon network closer to commercial deployment of reliable higher bandwidth speeds on the backbone to serve our customers’ needs – whether it’s voice, video or data,” claimed Mark Wegleitner, the company’s senior vice-president of technology.
However, Verizon’s penchant for announcing technology trials, rather than contract awards, probably indicates that the technology isn’t quite ready to be implemented commercially in a cost-effective way. FibreSystems Europe spoke to leading vendors of optical networking gear about the key challenges that still need to be overcome before 100 Gbit/s networks can be deployed on a grand scale.
How, why and when?
Verizon and other telecoms firms say the initial market pull for 100 Gbit/s will come from IP router ports because of their bandwidth utilization efficiency for packet flows. The other important driver towards 100 Gbit/s is, of course, fibre exhaust as cash-strapped carriers need to upgrade capacity on congested routes without laying new fibre. As always, cost is the underlying motivation as operators seek to lower the transport cost per bit on their backbone networks. “Although 100G might not be the most cost-efficient on day one, economy of scale will definitely reduce network costs,” said Ronan Mikdashi, ECI Telecom’s senior product marketing manager.
Using conservative estimates of traffic growth over the next few years, operators predict that 100 Gbit/s wavelengths will be required in volume in 2012, primarily in long-haul applications. The most urgent requirement for 100 Gbit/s has come from Verizon, which says it hopes to install 100 Gbit/s links as early as 2009. However, most systems vendors contacted by FibreSystems Europe say they expect first products to reach the market place around 2010. “Based on our research, we’re confident that 2010 will be the year when both standards and technologies will come together to create a boom in the 100G sector,” concurred Sterling Perrin, senior analyst with market-research firm Heavy Reading.
There are three organizations currently involved in defining the standards for 100 Gbit/s transmission: the Institute of Electrical & Electronics Engineers (IEEE), the International Telecommunication Union (ITU) and the Optical Internetworking Forum (OIF). The IEEE 802.3ba Task Force, which is on track to complete the 100 Gigabit Ethernet (100 GbE) standard in 2010, will hammer out physical layer specifications for the client side – how to connect racks of equipment over short distances from shelf to shelf or room to room. Meanwhile, the ITU is working on defining the transport mechanism for 100 Gbit/s over long-haul networks by extending the G.872 standard for the Optical Transport Hierarchy (OTH) to ODU4. The OIF has a wide-ranging brief covering 100 Gbit/s electrical interfaces, implementation agreements for transmit and receive modules and, most recently, has set up a group to investigate forward error correction schemes for 100 Gbit/s.
While most operators and vendors agree on the “why” and “when” of 100 Gbit/s networking, there is still a lack of consensus on exactly “how”. The important thing to note about 100 GbE is that it is a service interface only, and does not specify how that service should be carried over the network backbone. Similarly, OTH is essentially a network management protocol that packs 100 GbE and smaller tributaries into a transport “container”, and says nothing about the transport medium. Although the OIF’s work aims to provide some answers on how to implement 100 Gbit/s at the optical layer, so far it has been left up to equipment vendors and their carrier customers to decide the best way forward.
A recent field trial of 100 GbE illustrates this point. At the NXTcomm08 show in June, systems vendor Infinera teamed up with test equipment vendor Ixia and US carrier XO Communications to demonstrate the first ever transmission of a pre-standard 100 GbE signal. Ixia’s test set generated the 100 GbE signal and handed it off to a 100 GbE interface on the Infinera DTN optical system at the vendor’s booth in Las Vegas, which was then transported through XO’s long-haul network to Los Angeles and back again. One industry analyst called this the “fake 100G demo” because Infinera used 10 wavelengths, each running at 10 Gbit/s, to transmit the 100 GbE signal.
To be fair to Infinera, however, many network equipment manufacturers say they will use 10 x 10 Gbit/s as the first stage of rolling out 100 GbE. But while this approach works, and is relatively easy to implement using established 10 Gbit/s technology, it isn’t going to provide a viable solution for operators with capacity exhaust on their fibres. Those carriers need a way to transport native 100 Gbit/s traffic over long-haul networks, which ideally would be transmitted over a single wavelength (or possibly a closely spaced group of wavelengths) using a modulation scheme that is compatible with the 50 GHz channel spacings employed in reconfigurable optical add–drop multiplexer (ROADM) networks.
Unfortunately, the physics of high-speed signals in optical fibres turns 100 Gbit/s into a tricky proposition. Optical transmission was hard enough at 40 Gbit/s, when increased sensitivity to fibre impairments such as chromatic dispersion and polarization mode dispersion (PMD) forced the industry to abandon its beloved “simple blinking light” coding in favour of more complex modulation schemes based on phase rather than amplitude changes in the optical signal. At 100 Gbit/s the transmission impairments become at least 2.5 times harder to compensate, and may require new, more complicated modulation formats.
The roll out of 40 Gbit/s networking progressed through several different modulation schemes as the industry tried to find the optimum solution. First was carrier-suppressed return-to-zero (CS-RZ), then optical duobinary (OBD), and most recently differential phase-shift keying (DPSK). Another option, pioneered by Nortel, is dual-polarization quadrature phase-shift keying (DP-QPSK), which encodes two bits of data per bit of information transmitted on each of two polarizations.
Unlike the other modulation schemes, which maintain a symbol rate (the modulation speed of light on the fibre) equal to the true data rate, DP-QPSK reduces the symbol rate by a factor of four. Thus Nortel’s 40 Gbit/s optical system has a symbol rate of 10 Gbaud, allowing it to run over networks originally engineered for 10 Gbit/s. This feature has proved popular with operators, with Nortel announcing 22 customer wins to date, including tier-one carriers like Bell Canada. However, the downside is additional cost and complexity – the DP-QPSK transmit and receive components occupy a full line card. The technology also took Nortel six years to develop, according to Kim Roberts, the company’s director of optics research.
Lessons from 40G
In terms of 40 Gbit/s ports shipped, the systems vendors – notably Huawei and Nortel – have grabbed the largest market share, thanks to their internal development programmes, while the remainder of the 40 Gbit/s module market went to two start-ups – Mintera and StrataLight. In an environment where the technology is hard to do, R&D dollars were in short supply and the goal posts were constantly being moved, the business case for developing 40 Gbit/s modules simply didn’t stack up for established components vendors.
The upshot of all this is a 40 Gbit/s market place where component costs remain high and systems vendors have trouble finding second sources – a situation the industry is keen not to repeat at 100 Gbit/s. “The problem in the 40 Gbit/s market is that it is built on discrete components,” said Dave Skyes, vice-president of sales and marketing for 40 Gbit/s components vendor StrataLight. “As an industry we owe it to ourselves to do a better job on 100 Gbit/s.”
The OIF became involved in the 100 Gbit/s standardization effort this summer in an effort to speed up development of 100 Gbit/s components. It plans to draw up implementation agreements for integrated transmit and receive modules based on DP-QPSK. With a choice of modulation techniques available, narrowing the field to one specific scheme should help lower development risks by giving the industry a common starting point for designing hardware. “When I ask components vendors what they think of [the OIF’s proposal] they say that’s great, now I know what to build,” said Dana Cooperson, vice-president, optical networks at research firm Ovum. The OIF’s work is essential if the industry is to avoid the delays and costs that plagued 40 Gbit/s development, she says.
However, the optical components industry still has to overcome the fact that various technology pieces are scattered to the four winds. “If you look at the companies with the pieces, there’s nobody sitting there with a good electronics and optics play at 100G,” said Sinclair Vass, EMEA director for JDSU. Components vendors are going to need to collaborate to get the job done, he adds. JDSU is currently gathering the building blocks for 100 Gbit/s, which include tunable lasers, modulators, PMD compensation and polarization control. The optics vendor will probably fill any gaps in its portfolio by partnering with a start-up, as it did with Mintera on 40 Gbit/s modules, rather than acquiring the technology, Vass notes.
Proprietary solutions proliferate
With the OIF not specifying a timescale for completion of its work, it’s likely to be some time before integrated components for DP-QPSK become generally available. In the meantime, vendors with internally developed 100 Gbit/s systems solutions look set to reach the market first. Nortel’s first-generation 100 Gbit/s technology, expected to be commercially available in 2009, will use two optical carriers (wavelengths) spaced 20 GHz apart, with DP-QPSK encoding on each carrier to provide a symbol rate on the fibre of 12.5 Gbaud. Keeping the optical carriers just 20 GHz apart allows them to behave as one wavelength and fit nicely through the filter window of up to 12 concatenated ROADMs.
“The performance [of Nortel’s first-generation 100 Gbit/s solution] is not as good as at 40 Gbit/s – it only gives us 1000 km reach,” explained Nortel’s Kim Roberts. “The generation after that, which is what I’m working on, needs to overlay 10 and 40 Gbit/s networks and have greater than 2000 km reach.”
Nortel is not the only network equipment vendor pursuing an independent approach to 100 Gbit/s. Huawei demonstrated a 100 Gbit/s prototype this summer, although it did not specify which modulation format was used. For long-haul 40 Gbit/s transmission, the Chinese vendor favours a proprietary format called eDQPSK, which is a type of chirped return-to-zero DQPSK that has greater tolerance of nonlinear effects in the fibre than regular DQPSK.
In addition, US start-up OpVista, which uses a technology called Dense Multi-Carrier (DMC), demonstrated its 100 Gbit/s equipment at NXTcomm in June. Inside the company’s CX8 optical system, light from a single laser source is put through a “multicarrier generator” to produce five closely spaced optical carriers, onto which five separate streams of 20 Gbit/s data are then modulated using QPSK. Recalling that QPSK transmits two bits of data per symbol, the symbol rate on the fibre is 10 Gbaud.
Whatever solution the industry comes up with, equipment vendors say it is vital that the cost of high-speed optical components comes down. “We’re going to be heavily reliant on what the components vendors come up with, otherwise it [100G] will be too expensive,” said Vinay Rathore, senior director of marketing, EMEA, at Ciena. “If we come in at 60% of the cost of 10 x 10 Gbit/s then we’ve hit the mark.”
If the OIF’s choice of DP-QPSK is widely adopted by components and systems vendors, it will help drive down prices by boosting production volumes. Ultimately, however, whether it’s possible for the industry to standardize on a particular modulation scheme at 100 Gbit/s depends on cost. “If DP-QPSK comes in more expensive than other modulation schemes, it forces the scenario where you choose the scheme depending on the application,” said Kim Papakos of US systems vendor Tellabs.
Naturally, service providers are hoping that future 100 Gbit/s solutions will be priced more competitively than 40 Gbit/s equipment is now. Some experience with advanced modulation formats and components can be re-used in the jump from 40 to 100 Gbit/s, which should help speed up the introduction. “The leap to 100G from 40G is not as big of a leap as it was to 40G from 10G,” commented Heavy Reading’s Perrin. Even so, he says that’s no guarantee that 100 Gbit/s prices will come in at the 2.5 times price parity that operators desire when equipment becomes available. “It may not be 2010 when the market hits the price points that service providers want,” Perrin concluded.