Recently in Editor's blog Category

The blogosphere is buzzing today with the news that KPN and Reggefiber plan to spend €6–7 billion in an ambitious project to roll out fibre-to-the-home (FTTH) across the whole of The Netherlands over the next 5–7 years.

The Netherlands is currently 12th in the global FTTH rankings, which are compiled every six months by the three FTTH Councils, with fibre reaching 1.4% of the population. This development will certainly move the country up the list.

Of course, it helps that the Netherlands is a relatively small country, with a high population density — the highest in Europe, not counting "island" states like Malta and Gibraltar. There are around 7 million homes to pass, which makes the cost per home around around €1000, and the carriers plan to foot the bill entirely by themselves, without government aid.

The original article breaking the news in Dutch paper Trouw is here, the headline of which translates rather ungracefully as "KPN and Reggefiber stabbing 6 billion in national glasnet".

Artist concept of interplanetary internet. Credit: NASA/JPL

Perhaps being a settler on Mars won't be so tough after all — at least you'll still be able to check your email, if research at NASA comes to fruition.

The US space agency has successfully tested the first deep-space communications network modelled on the internet. In a series of experiments that started last month, NASA engineers transmitted dozens of space images from Earth to a NASA science spacecraft located more than 20 million miles (32 million km) from Earth, and back again, using a network consisting of 10 nodes.

"This is the first step in creating a totally new space communications capability, an interplanetary Internet," said Adrian Hooke, team lead and manager of space-networking architecture, technology and standards at NASA Headquarters in Washington.

Vint Cerf, who is Google's vice-president and chief internet evangelist, was instrumental in creating the technology, called Disruption Tolerant Networking, or DTN. The DTN software protocol differs from the normal Transmission-Control Protocol/Internet Protocol (TCP/IP) communication suite because it does not assume a continuous end-to-end connection.

The interplanetary internet must be robust enough to withstand delays, disruptions and disconnections in space. For instance, the delay in sending or receiving data from Mars takes between three-and-a-half to 20 minutes at the speed of light. Longer communications blackouts can result when a spacecraft moves behind a planet, or during solar storms.

Unlike TCP/IP on Earth, DTN does not discard the data packets if a destination path can't be found. Instead, each network node keeps custody of the information as long as necessary until it can safely communicate with another node. This store-and-forward method means that information does not get lost when no immediate path to the destination exists. The information always gets to the end-user — eventually.

DTN will be a boon for space communications. "In space today, an operations team has to manually schedule each link and generate all the commands to specify which data to send, when to send it, and where to send it," said Leigh Torgerson, manager of the DTN Experiment Operations Center at JPL. "With standardized DTN, this can all be done automatically."

DARPA is very interested in DTN too, for military applications. And to my mind, this research sounds like it could be quite useful on Earth as well, especially in areas that currently suffer from poor or intermittent network connectivity.

Could Gemfire be the first casualty of the credit crunch? The integrated optics firm is reportedly running short of cash as investors get cold feet and customers push out their orders.

As a result it has abruptly closed down its facility in Livingston, Scotland, with the loss of 250 jobs, says this BBC report. (You may recall that the Livingston plant is the home of the former silica-on-silicon firm Kymata, which was sold to Alcatel Optronics, then to Avanex, and finally to Gemfire.)

The fiesty Scots don't appear to be very happy at this treatment, claiming that the Livingston outfit was the most profitable of the three Gemfire locations, and the order book was full, according to this article in Scottish daily The Herald.

However, UK labour laws apparently don't allow the option of putting the company on ice for two weeks, the tactic the company has employed at its two US locations to see it through the tight spot.

If — and I'd say that's a fairly big if in the current climate — Gemfire manages to conjure up additional funding, either in the form of a new investor or a bridge loan from customers, then it may try to reopen the Livingston factory.

In the midst of the turmoil in the banking industry, and the seemingly endless stream of bad news from Wall Street, it's easy to get stuck in a negative mindset. But a couple of articles that I saw recently highlight the fact that one man's bad fortune can be another man's opportunity.

For a kickoff, Paul Graham, essayist and co-founder of Viaweb, the first application service provider, writes about why to start a start-up in a bad economy. Many folk are predicting that this economic slowdown could be as bad as the seventies, he notes... that's when Apple and Microsoft were founded.

Investment guru George Gilder also provides some welcome relief from the doom and gloom with an article in Forbes in which he talks about an upcoming creativity boom. "The real source of all growth is human ingenuity and entrepreneurship, which often thrive in the worst of times," Gilder points out. Which company emerged victorious from the telecoms crash of the early 2000s? Google.

A few more things I saw at ECOC are worth noting, albeit belatedly, and a bunch of them were about 40 Gbit/s components.

In his market focus presentation, Kim Papakos from Tellabs highlighted single source components as a real problem for systems vendors wanting to build 40 and 100 Gbit/s networking gear. Ideally companies like Tellabs would like at least three sources on the market, so that if one supplier goes belly up, there are still multiple sources available. That requirement is passed down to them from their operator customers.

Loi Nguyen and Lian Zhao from Inphi

The lack of suitable off-the-shelf parts for 40 Gbit/s has often been cited as one of the factors that delayed the adoption of 40 Gbit/s networking. Nevertheless, it surprised me to find that there are still quite a lot of 40 Gbit/s components that are only made by one company, and some that simply aren't available at all.

A case in point, Inphi and Sierra Monolithics joined forces to create the industry's first reference design for 40 Gbit/s DQPSK (differential quadrature phase-shift keying) modulation. The booth demo showed interoperability of Inphi's 40 Gbit/s modulator driver with Sierra Monolithics' SerDes multiplexer/clock multiplier unit.

As I understand it, Inphi is the sole supplier of DQPSK modulator drivers, while Sierra Monolithics is currently the sole supplier of 40 Gbit/s SerDes. That's because high-speed, high-voltage chips are difficult to make in silicon, so vendors have to resort to other materials systems.

Inphi's DQPSK modulator driver for example, which has to output two streams of 20 Gbit/s at an amplitude of 8 V, is made out of gallium arsenide. (For comparison, networking chips in other roles typically require 1 V.) Meanwhile Sierra Monolithics makes its present generation of chips out of silicon germanium using IBM's 7HP process.

The supply situation will improve over time of course — CoreOptics announced a 40 Gbit/s SerDes chip earlier this year, and AMCC is possible also working on one. But right now the components supply chain isn't complete at 40 Gbit/s, let alone 100 Gbit/s.

In fact, Loi Nguyen, co-founder and vice president of technology for Inphi, told me: "Nobody has a TIA [transimpedance amplifier] for 40 Gbit/s DP-QPSK [dual-polarisation quadrature phase-shift keying]." The TIA sits on the recieve side in the equivalent position to the modulator driver on the transmit side, and amplifies the weak electrical signal coming in from an optical detector before passing it to the SerDes.

DP-QPSK is the modulation format selected by the Optical Internetworking Forum (OIF) for 100 Gbit/s components, which is supposed to hit the market in 2010, and yet it's still not possible to get standard parts for the previous speed incarnation. Clearly components vendors have a lot of work to do.

Despite the prevailing gloom in the financial sector, vendors exhibiting at the European Conference on Optical Communications (ECOC) were generally upbeat, with one notable exception — Avanex.

Well, a possible reason for the glum demeanour of the employees on its booth just came to light. According to an 8K form filed last week Avanex plans to lay off about 8% of its work force, or 47 employees by the end of this month. The company also plans to close its Melbourne, Florida facility and transfer the product lines, inventory, and fixed assets to its France or China offices. What's more, the executive officers have voluntarily agreed to a 10% reduction in their salaries beginning October 2008.

Avanex seems to be tightening its belt in anticipation of reduced sales in the later part of this year. Although the company posted 5% sequential growth for Q408, which ended on 30 June, it issued guidance for a substantial sequential revenue decline for the following quarter. The company has yet to find a permanent replacement for former CEO Jo Major, who was axed in August, and it's stock has reached an all-time low.

Chaos was the order of the day at the Brussels Eurostar terminal on Thursday, where the reduced service after the fire the previous week left hundreds of travellers queueing to get their tickets validated for one of the few remaining trains. Looking on the bright side, however, if I'd been travelling by plane, I probably wouldn't have got home at all due to an air traffic control failure.

Here are some less stressful memories from the event:

Brussels Expo Hall 10

The Atomium

Beauty parade

Ready for action

FibreSystems prize draw

Mattias Persson & Ted Takeuchi

ECOC exhibition

Components vendors are out to prove that Infinera is not the only game in town when it comes to photonic integrated circuits (PICs). At the European Conference on Optical Communications (ECOC) there are a couple of demos worth pointing out.

First up, tunable laser maker Santur is showing a 100 Gbit/s PIC made out of indium phosphide. The device is a variation on the company's tunable laser technology, which contains an array of 10 lasers with a MEMS device to select the output from one of them. In the 100 Gbit/s PIC, the MEMS device has been replaced by a passive multiplexer, which combines the light from all the 10 Gbit/s lasers (running at the same time) onto a single fibre.

The company claims that its device offers a simple solution for distances up to 10 km, in access networks and short reach interconnects. It could potentially be cheaper than the 4x25 Gbit/s or parallel optics schemes that are also being considered for 100 Gbit/s applications.

But Centre for Integrated Photonics (CIP), the former research activity of BT, has gone one better with a demonstration of a 32-channel PIC.

CIP's photonic integration technology, which was unveiled here at the show is based on hybrid integration platform called HyBoard. Indium phosphide active components are mounted onto silicon "daughterboards", which are then flip-chip mounted onto a silica-on-silicon motherboard containing optical waveguides to connect all the components. The platform is self-aligning, and has been designed so that it can be outsourced to a contract manufacturer if volumes require it.

CIP's demonstration, which contains continuous-wave (CW) lasers, showed how the channels could be turned on and off individually, how the power could be balanced across the channels, or tilted to suit the input to an optical amplifer. Inside the package are two monolithically-integrated arrays of 16 lasers each, mounted side-by-side on the HyBoard platform.

The PIC in its current form is intended for WDM-PON, where it could provide a colourless solution for customer premise equipment (CPE). In this application, an array of CW lasers located at the headend sends light down the fibre to the CPE, which contains a reflective SOA that puts the data onto the wavelength, and reflects it back up the fibre to the optical line terminal.

CIP says that the concept could be extended to create a 160 Gbit/s PIC, containing 16 channels at 10 Gbit/s, simply by adding modulators to the package.

Rod Alferness, chief scientist, Bell Labs.

Trying to predict the future is a good way to invite egg on face, as many people throughout the ages have discovered. But yet that's exactly what plenary speakers are expected to do when they take to the stage, and at the European Conference on Optical Communications (ECOC) in Brussels today it was the turn of Rod Alferness, chief scientist at Bell Labs.

Alferness had been asked to talk on the same topic at ECOC in Madrid 10 years ago, and before looking forward to the next 10 years, he took the opportunity to look back to see how his predictions had turned out.

Back then 1 Tbit/s had already been demonstrated, so transmission was essentially a solved problem. The real issue for the optical networking industry at the time was how to "manage and tame" all this bandwidth. Mesh and reconfigurable networks based on ROADMs looked like the optimum solution, and in fact that has turned out to be the case, even if it has taken a little longer to arrive than expected (no rotten tomatoes so far).

But going forward, increasing transmission capacity will not be so easy. The belief that fibre has unlimited bandwidth has been overturned. "I think that in the next 10 years we are going to be very challenged to find the bandwidth that society demands of us in a cost-effective way," Alferness said.

The fundamental drivers for bandwidth growth aren't going away, the opposite in fact, as new applications like high-definition TV, video on demand, and 3D video-conferencing could stimulate demand in an unprecedented way, particularly if the bottleneck in the access network is tackled.

While there are differences of opinion on how much bandwidth will grow, even the most conservative estimates suggest that some significant advances in network technology will be required. Alferness presented a chart showing how the capacity demonstrated by "hero experiments" is increasing far more slowly than before. "In the past we would have said a breakthrough was required," he observed.

Although it's impossible to predict what that breakthrough could be, advanced modulation formats, which make more efficient use of fibre spectrum, are definitely part of the picture, and so is photonic integration. "It is absolutely clear in my mind that substantial integration is going to be critical [to high-capacity transmission]," he said.

Ultimately Alferness believes that optical components will be integrated with electronics. "I think 10 years ago we thought optics had all the answers, now I think optics and electronics are going to have to work together," Alferness proposed.

The marriage of optics and electronics could take place in either indium phosphide materials and silicon photonics, which could turn out to be very interesting technology race. However, the horizontal nature of the business model will complicate the investment case for new components technologies, he warns.

The next 10 years

Here's Alferness' take on other trends for the industry over the next 10 years:

• On the overall network: It will be one network. Increasingly the data layer and the optical layer are converging. If we have intelligent routers at the edge of the network, the efficiency of that network is going to be improved.
• On green issues: Energy and carbon footprint will be critically important as we design new optical systems, and it will influence the decisions that are made on technology, particularly decisions about architecture.

• On WDM-PON: WDM-PON will happen at least to enterprises if not to homes.
• On wireless: Wireless needs an interconnection network, which is provided by optical. Fibre networks that go out to homes will also go out to wireless base stations to provide backhaul.
• On multiple antennas: Base stations should talk to each other. 3–5 base stations could provide communications to a single user, by strengthening the signal instead of interfering with one another. But they can only do that if there is an ubiquitous low-cost interconnect between base stations, and that's optical, obviously.

• On network storage: Increasingly we will put storage in the network, closer to the end user to reduce transmission requirements. But its' a wild card: more storage in the network means more content will available to users, which will help to drive traffic.
• On telepresence: We need to offer people the opportunity to really get together without travelling, whether that's telepresence or something else. We have flattened the world, and built a global community, and people will be looking for more personal ways to communicate.

• On in-home networks: We will see wireless networks based on femtocell base stations. Initially the broadband connection in your home will backhaul the traffic, in future, femtocells may also serve traffic coming down your road.

• On quantum computing: Is it the Google of the next decade? Quantum computing enables massively parallel operations. It could perform searches several orders of magnitude faster than today's searches, and in 10 years we're going to need something that does searches that much faster.

Most headlines involving Google seem to involve "world domination", and yesterday's article from The Times online stuck to the theme, as it reports that Google may take its battle for global domination to the high seas with the launch of its own "computer navy".

In fact, Google has applied some lateral thinking to the problem of how to power its energy-hungry data centres. It's latest idea is to create floating data centres located up to 7 miles (11 km) from shore, in 50 to 70 m of water. If perfected, this approach could be used to build 40 mW data centres powered and cooled by the ocean.