Under the sea with 40G

The submarine cable boom is back with a vengeance. In recent months, Asia has seen several new undersea systems kick off, such as Tata's TGN-IA, which links Hong Kong, Singapore, Tokyo, Vietnam and the Philippines with initial design capacity of 3.8 Tbps and the Asia America Gateway cable, linking southeast Asia to the US. Meanwhile, Pipe Networks' PPC-1 cable, which links Sydney, Guam, Papua New Guinea, Tokyo and San Jose, was set to launch this month as we went to press.

That's just in the Pacific.

Research firm TeleGeography estimates that by the time 2009 is over (if all goes according to plan) 16 new undersea cables will be in place. That's more than the number of submarine systems built in 2001, the height of the subsea bandwidth bubble. And there's more where that came from - TeleGeography also says around $2.4 billion worth of new submarine cable projects are scheduled to be finished in the next two years.

The motivation behind the push for all this fresh subsea capacity isn't a mystery. Some more numbers from TeleGeography: international traffic growth accelerated to 79% this year, up from 61% in 2008. Tellingly, much of that is coming from emerging markets like Eastern Europe, South Asia and the Middle East, where traffic growth this year alone has topped well over 100%. Break it down by country and you find that Vietnam's bandwidth usage is expected to grow from 22.8 Gbps in 2008 to grow 13 times to 302.4 Gbps in 2013. Thailand will see a tenfold growth in bandwidth requirements in the same time frame.

"We're seeing tremendous demand that's not always coming from the usual places," says Pacnet CEO Bill Barney. "We're looking at 3G, Chinese broadband deployment, video, cable TV on broadband in China, Indian consumer connectivity going bananas, also Indonesia and Thailand."

Little wonder network operators have added 9.4 Tbps of new capacity this year, TeleGeography says.

Naturally, keeping up with that hunger for bandwidth isn't just a matter of building new cable systems but also upgrading capacity in existing links.

Consequently, carriers are entertaining various options to accomplish that. One option gaining traction is upgrading 10G wavelengths to 40G.

40G wave technology has been around for a couple of years and has seen some deployment in terrestrial optical networks. But various trials aside, it has yet to really gain traction in the subsea market, primarily due to the wet nature of the business. As most of the network plant is underwater, boosting wavelength bit rates has to be done using SLTE (submarine line terminal equipment) gear plugged into the dry plant at either end of a given link. A number of carriers have already done this to upgrade 2.5G waves to 10G - Pacnet did it this past May, for example, using SLTE cards from XTera to upgrade its EAC-C2C system.

Taking that to 40G has been problematic, partially in terms of the economics (as it has been generally cheaper for carriers to deploy four 10G cards than one 40G card), but also in terms of reach - 40G has to not only be able to cover relatively short links of a few hundred kilometers, but also transoceanic links that can span 9,000 km or more.

But that could be changing soon. Vendors like Nortel Networks (its bankruptcy proceedings notwithstanding), Infinera and Huawei Technologies are bringing their terrestrial optical technology to the SLTE space, challenging the big subsea suppliers (Alcatel-Lucent, Fujitsu, NEC and Tyco) with SLTE gear that they say can help 40G go the distance.

SLTE: plug it in

Nortel's 40G SLTE solution has drawn considerable attention in the last couple of months after the company publicized a trial of the solution with subsea cable operator Southern Cross Cables Network.

"Southern Cross gave us a route of around 4,500 km from Hawaii to the US west coast, Segment D on their route," says Anthony McLachlan, VP of carrier networks for Nortel Asia. "We turned that up in a matter of days. We didn't touch the wet plant, we just plugged in our OME 6500s with 40G cards, and it worked."

While Southern Cross was impressed, McLachlan continues, "They said, 'if I have to light up one segment at 40G, I have to be able to do the lot, and you'll need to run 40G on Segment C, which is 8,000 km between Hawaii and Auckland'. So we did that too."

The secret to Nortel's ability to send 40G at that reach is the use of DPBPSK (Dual Polarization Binary Phase Shift Keying) modulation, which gives it a 3dB increase in reach and 6dB more resilience to phase distortion, enabling longer distances than before.

Infinera, meanwhile, is making its subsea push via its DTN solution based on the company's 100-Gbps photonic integrated circuit (PIC) technology, which integrates several hundred components and supporting 10 x 10G channels (with 25-MHz channel spacing) in a single optical chip on a submarine line module (SLM) that can reach distances of 6,500 km.

"These 100-Gbps submarine line modules map any client service - 2.5G, 10G, 40G and in future 100G - over 100Gbps of PIC-based WDM capacity," explains Serge Melle, Infinera's VP of technical marketing. "The PICs carry information over ten wavelengths operating at 10 Gbps each, but users can use the bandwidth as if it was one combined pool of 100G, using as little or as much of that as they need for a given service."

On the downside - at least for now - that means Infinera's 40G solution is still essentially a 4x10G solution, notes Ovum analyst Ron Kline.

"Infinera's approach requires the use of four wavelengths, compared to other vendors that use a single wavelength to provide 20G or 40G services," he said in a research note. "This puts Infinera at a disadvantage over some of its rivals because four wavelengths must be used for one 40G circuit. This disadvantage will last until its own 40G PIC technology (expected to be released in 2H09) is made generally available and adapted for this application."

That said, Kline adds that Infinera does offer other potential benefits. "Replacing existing SLTE terminals with DTN also gives subsea cable operators the rapid provisioning features that are available with Infinera's terrestrial systems. Operators that deploy both terrestrial and SLTE products from Infinera stand to benefit from much reduced optical-electrical-optical signal conversion costs at landing stations."

Cable diversity

The catch - and you knew there would be one - is that upgrading subsea cable capacity is not quite as simple as plugging in a 40G SLTE card, according to the old-school submarine cable suppliers that built the systems currently in the water.

For example, says Olivier Gatutheron, R&D technical director for Alcatel-Lucent's submarine network business, some subsea fiber systems are simply not designed to support waves faster than 10G.

"For example, we have deployed long-haul systems using DPSK technology and specific fiber - dispersion managed fiber - which provides capacity of 1.6 Tbps on one fiber for 9,000 km, but this is not upgradeable to 40G because of physical limitations," Gatutheron told Telecom Asia. "This is because the repeaters are spaced far apart in order to make the wet plant cost-effective."

Gatutheron says that shorter links could be upgraded to 40G with SLTE cards, but even then, it still depends on the link characteristics.

"There are many different types of cables in the water and many different generations," he says. "If you look at old systems with 2.5G WDM, you could upgrade those with SLTE that's not very state-of-the-art, because it's an old system so you can use low-cost simple terrestrial technology on it. What's more difficult is very long systems and newer links. Again, repeater spacing is quite large and variable, it's not like terrestrial where the position of the optical amplifier is fixed."

Another issue - assuming that operators upgrade some but not all of their existing waves to 40G - is the technical challenge in running waves running at different speeds on the same fiber, Gatutheron adds. "When you mix bit rates, you have interaction between wavelengths that causes degradation of the 40G channel, so you have to be careful with what modulation format you use on 40G waves."

Gatutheron also notes that SLTE-only solutions have to do more than just send a 40G wave far enough to work on long subsea links. "It also has to provide a sufficient margin for fiber aging, cable repair and performance situations."

Case by case

Assuming the technological issues can be mitigated, there is still the question of just how much demand there is for subsea 40G, regardless of the internet traffic growth hockey stick.

Some players, like Alcatel-Lucent, say subsea 40G demand will only grow once 40G takes off in the terrestrial networks.

"Terrestrial backhaul is a key driver for 40G in the submarine network," Gatutheron insists. "At the moment there's not very much deployment of 40G in the terrestrial side, so there is not strong demand today. The need for 40G in submarine will come when there's more 40G in terrestrial."

Another factor is that subsea systems in Asia are sitting tons of unlit capacity, according to TeleGeography (see chart page 23), with around 85% of total capacity unused. Pacnet - whose EAC-C2C system accounts for around half of the region's subsea total potential capacity - is planning to light up almost another terabyte later this year, says Barney.

"That's three times the size of any upgrade we've done before, and about 25% of it already sold," he said. But while Pacnet is trialing 40G, it has no concrete plans to adopt it any time soon.

Meanwhile, neither the amount of unlit capacity nor the promise of 40G upgrades at the subsea terminals is stopping carriers from investing in new systems, or even adding links to existing systems - even the ones actively trialing 40G.

In September, for example, Southern Cross' director of sales and marketing, Ross Pfeffer, told Telecom Asia that the operator is formally looking into building a new cable from New Zealand to the US, citing the upcoming surge in traffic that NBN plans in Australia and New Zealand are expected to generate.

And while SLTE solutions are being touted as cheaper options to building new systems, TeleGeography analyst Alan Mauldin points out that many of the new systems are simpler and cheaper to build. "Most of these new cables cover shorter distances and employ simpler designs than their predecessors, helping to keep costs in check." For reference, he adds, compare the $2.6 billion being spent on cable deployments this year - which, remember, works out to more systems than the ones built during 2001 bubble - to the $13.5 billion invested in new systems during that bubble.

Of course, one of the issues informing these new systems is not just raw capacity but also latency. As internet traffic - much of which is increasingly time-sensitive, like video and VoIP - grows in certain areas, the engineering challenge isn't just to get it there in the fastest time possible, but also via the most direct path. Which is why one of the hot selling points of the new TGN-IA systems was direct fiber paths within Asia that enable reduced latency times (Tata claimed that its Singapore-Tokyo link sports an RTD of 63ms).

Maximizing cable life

Still, even if all that means the 40G SLTE market will be limited to case-by-case uptake, it's still a market that Ovum's Kline says will grow from $843 million in 2Q09 to $2.2 billion by 2014 (though how much of that will be 10G, 40G or even 100G isn't clear).

In any case, it's a real market that SLTE vendors are more than willing to serve, says Nortel's McLachlan.

"There's a need to build a lot of latent capacity on these systems that operators can sell," he says. "They're still doing 15-year IRUs on ten-year-old cables, so they're looking to extend the life out of them, and to do it as quickly as possible. If I can double or quadruple the capacity on a particular link and do it fast, that's significant revenue potential for that asset they've got in the ground. What we're saying to the existing cable operators is that they have an opportunity to add incremental capacity to a sunk investment, which allows them to revalue their assets and drive more revenue from it." 

Up next: IP/optical convergence

Much of the innovation for subsea optical technology is often driven from the terrestrial side of the optical business. 40G is one emerging example. The next big innovation could be the blending of the IP and optical layers into one cost-optimized layer that could not only save carriers serious capex/opex savings, but also improve latency.

That's the pitch behind Alcatel-Lucent's "converged backbone transformation" (CBT) solution. And while it may sound like another IP-over-DWDM solution - and indeed, IPoDWDM is the starting point for CBT - it's "much more than that", says Alberto Valsecchi, Alcatel-Lucent's optics marketing VP.

"We are talking about data-plane, control plane and management plane integration between the IP and optical layers," he says.

Valsecchi touts Alcatel-Lucent's optical and router expertise as an advantage over vendors focused mainly on one or the other - and partnerships don't count, he said in an apparent reference to the partnership formed by Nokia Siemens Networks and Juniper Networks in June this year to develop IPoDWDM solutions.

"There are some things that can't be done through partnerships and we believe and core/optical convergence is one of them," he says. "It's much harder to offer this solution through partnerships in terms of aligning road maps, interoperability testing. With multiple vendors, it gets tricky."

CBT addresses one of the chief downsides of IPoDWDM: inefficiency. IPoDWDM maps one router port to each optical wave, which means the wave becomes underutilized during off-peak hours.

Alcatel-Lucent intends to provide traffic grooming options from the wave level (i.e. IPoDWDM) to the sub-port level using ODUflex technology, an emerging ITU standard due for completion next year, which provides higher granularity by enabling VLANs or pseudowires within a port to be logically or virtually mapped to the same wave.

Result: carriers can maximize capacity without spending more money on extra core routers, and yield capex savings of "at least 30%, in addition to savings in power, space and operational complexity".