Entry No.06t

IT Writers Awards

Speed differences between corporate LANs and WANs have long been a problem for many companies, who have been forced to develop highly decentralised network infrastructures to compensate for WAN technologies’ chronic inability to shuttle data between sites at anything even remotely resembling LAN speeds.

But hang onto your hat: the exploding reach of fibre-optic networks, turbo-charged by the coming explosion of optical networking technology, is going to radically change the ground rules of wide-area networking as they get access to WAN bandwidth beyond their wildest dreams.

Beating the bandwidth crunch
Whereas conventional networks transmit data using electrons, optical networks rely on fibre-optic cabling to carry data using light pulses instead. Fibre-optic networks are already well-understood for their high capacity and immunity to electrical interference, which has made them a favourite of many companies as internal network backbones.

Over longer distances, expensive fibre has long remained the exclusive domain of cashed-up telcos and billion-dollar international consortia. But with factories the world over now producing thousands of kilometres of fibre every hour to meet escalating demand, however, its cost is finally declining rapidly.

But laying fibre requires owning or leasing right of way, and it’s expensive and time-consuming to buy access to -- then dig up and refill -- large expanses of earth. Although utility companies such as United Energy Telecommunications and PowerTel have avoided this problem by running fibre through their existing rights of way, growing demand for bandwidth has pushed every carrier to consider other ways to increase their available bandwidth.

And rather than simply laying more fibre, in recent years telcos have warmed to new technologies that allow multiple data streams to be squeezed over existing fibres. The first of these was Time Division Multiplexing (TDM), which carries multiple signals over a single cable by breaking up multiple signals and transmitting them as a continuous stream of information that is reassembled at the receiving end.

Wavelength Division Multiplexing (WDM) takes a different approach, using two, four or eight separate beams of light to carry multiple streams of data at once. This technique is more appropriate than TDM for carrying data, since TDM relies on stuffing data into the natural pauses in voice conversations. WDM, on the other hand, allows transmission of multiple, continuous streams of data.

WDM technology is now widely used across most carrier networks around the world, providing instant bandwidth increases of up to eight times over a single fibre. This has allowed carriers to rapidly increase their capacity to carry conventional voice calls at relatively low capital cost -- an increase that is compounded when WDM is applied to the multi-strand bundles of fibre in telecommunications trunk lines.

Full-throttle bandwidth
But even WDM is starting to show signs of strain as volumes of Internet traffic steadily increase. These volumes will skyrocket later in the year as multi-megabit xDSL, cable modems and satellite connections become the preferred method for businesses and consumers alike to get online. Their spread will enable pervasive telecommuting with data access at speeds perceptually comparable to those of a LAN, but carriers need to bulk up their networks to cope.

This boost will come from DWDM (Dense WDM) technology, which works along similar lines as WDM but uses multiple lasers to provide 16, 32, 64 or more simultaneous data streams, each using different wavelengths (colours of light). Current DWDM technology can transmit 2.5 gigabits per second (Gbps) per wavelength, or lambda, but by the end of this year this will have increased to 10 Gbps per lambda. Speeds of 40 Gbps per lambda are expected within a few years, at which point systems carrying hundreds of lambdas per fibre should be generally available.

DWDM has a long way to go before it maxes out, says Bob Martin, US-based chief technical officer with Lucent Technologies. "We’ve run 3.2 terabits [Tbps] over 300 kilometres over a single fibre [in tests]," he says, "which is roughly equivalent to all of the conversations taking place in the world at any one time. We’ve run 1022 wavelengths on a single fibre, and I see no reason why that shouldn’t scale to 5000. We think in the not-so-distant future we could get on the order of 100 Tbps over a single fibre."

While such speeds are enough to make many a carrier drool with delight, their introduction will be limited by some practical complications. Namely, variations in the quality of fibre – particularly that which has been installed for several years – may force some carriers to install newer fibre cabling with far lower dispersion rates that support such high-speed transmission. There will also be delays as both standards bodies and equipment manufacturers come to grips with the innovative technologies needed to make virtually unlimited bandwidth a reality.

The optical gold rush
The worldwide market for optical networking equipment is exploding; analyst firm Dell’Oro Group, for one, recently reported that the optical transport market grew 22 percent in the last quarter of 1999 to be worth $US4.9 billion – a run rate that should put this year’s market well over 1999’s $US15.3 billion.

DWDM equipment last year accounted for some $3.8 billion of this, with the rest accounted for by more-mature SONET/SDH (Synchronous Optical Network/Synchronous Digital Hierarchy). SONET (used in the US) and SDH (used in Australia and elsewhere) split raw fibre into manageable chunks of bandwidth that can be onsold as telecommunications services. These networks have traditionally provided high-speed connections between local and interstate telecommunications exchanges, which have been terminated at customer premises over the local cooper loop or, occasionally, local fibre connections.

Melbourne-based Haliplex Communication Systems, a 13-year-old networking equipment developer, is elbowing its way into the global optical networking market with two optical networking multiplexers that provide direct access to SONET/SDH optical networks already widely used by most carriers.

But by distributing multiple smaller interfaces across their SDH network, Haliplex sales and marketing director Alan Kepper says carriers will be more able to scale xDSL and other broadband services to match market adoption without requiring the expense of full-on telecommunications exchanges. He believes Australian carriers will instead extend their SDH networks straight into local neighbourhoods using unobtrusive, traffic signal switch-sized multiplexers that can be installed in parking lots, on street corners, or anywhere else they’re needed.

"It’s a fundamental mathematical equation," Kepper explains. "You’ve got more and more high-speed access, and there’s technically no way of reticulating the bandwidth of everyone except over optical. When you’ve got all this access with people hooking up with [xDSL and] fibre, the core has got to be incredibly thick."

Although it’s seeking funding to take its technology into the world market, Haliplex will be stepping into the path of hungry competitors in an optical networking market that is expected to grow to more than $US13 billion in North America alone by 2003. Potential rivals include Nortel Networks, Lucent Technologies, Alcatel, Fujitsu, Marconi, and Ciena – and all are fighting startups such as Sycamore, Redback and Chromatis to dominate.

Not surprisingly, this mad dash has been fuelled by a number of big-budget takeovers: market leader Nortel, for one, has recently paid $US8 billion to buy laser and filter maker CoreTek and optical networking startups Xros and Qtera -- neither of which has even started shipping products. The company recently announced it will invest some $US660 million into optical networking to triple its production capabilities in Canada, the UK and US, and last month created a completely new business division, called High Performance Optical Components Solutions, to focus its optical networking development efforts.

Competitors have been chasing the market equally hard. Cisco, for one, has shelled out over $US10 billion in the last year to purchase optical networking firms Cerent and Monterey Networks, and to acquire the optical networking-related assets of long-haul DWDM specialist Pirelli. Last November, Redback Networks bought nascent Siara Systems for $US4.3 billion from its $US6 billion IPO last year. And in March, Lucent bought out optical networking startup Ignitus Communications for an undisclosed amount.

Working in fierce competition, these companies – and others – are rapidly advancing the state-of-the-art. Later this year, for example, DWDM will be challenged by a proprietary standard from SilkRoad called SilkRoad Refractive Synchronization Communication (SRSC), which is said to offer up to 6 Tbps of transmission at a lower cost than DWDM because it uses a single laser beam.

New photonic switches from companies such as Xros and HP spinoff Agilent will provide a major boost for optical networking, since both solve a long-standing problem with optical networks – specifically, the need to convert photons received over the fibre cable into electronic signals, switch the traffic, and then convert the electrons back into photons for retransmission to their destination. This is because current equipment cannot pull the destination information it needs straight out of the light stream.

Instead, Xros, Agilent and the inevitable stream of future competitors have created switching platforms that use arrays of microscopic mirrors to switch the light beams without having to convert them to electrical impulses. This will remove the overhead of photon-to-electron-to-photon conversion and pave the way for stunningly fast network switches.

Not one to be left behind, Intel has also made moves into optical networking, in March spending $US1.25 billion to purchase Danish optical networking chipmaker Giga A/S. But Intel will be following the lead of chip maker Broadcom, which last month lit a fire under equipment makers with the launch of a single $US88 chip providing 10 Gbps optical transmission.

Broadcom’s chip will no doubt dovetail with current efforts by the IEEE to produce a 10-Gigabit Ethernet (10 GE) standard, IEEE 802.3ae, which will provide a standardised method for corporates and service providers to interface with massive optical networks. Pre-standard 10 GE equipment is expected from Cabletron and others late this year, and Dell’Oro recently predicted the market will pass $US1 billion by 2004.

Fibre, fibre everywhere
Although the promise of significant bandwidth increases will excite many of the world’s largest consumers of bandwidth, the value of that bandwidth will only be realised as practical applications begin to surface. One such project is CANARIE (Canadian Network for Advancement of Research, Industry and Education), a $15 million Canadian government-sponsored optical network that uses 16-lambda Nortel Networks equipment to span the thousands of kilometres from Vancouver in the west to Newfoundland in the country’s east.

Bell Canada has the right to half of the CANARIE network to carry its voice calls, and only two of the eight remaining data wavelengths have been lit yet. Yet even so, senior director of network projects Bill St. Arnaud says the abundance of bandwidth – fuelled by government legislation allowing competing carriers to inexpensively string fibre-optic cabling between above-ground electricity poles -- has already driven major changes in WAN strategies.

"DWDM technology is going to fundamentally change the customer-carrier relationship," he explains. "Right now, carriers build networks out to customers – but with optical technology in conjunction with low-cost dark fibre, customers will build optical networks of great capacity out to the carriers. Many school boards are running dark fibre to every school, and whole bunches of institutions are getting together to own fibre. It’s in the wide area, but using LAN technology such as Gigabit Ethernet straight out of the box. We’ve peeled back the covers and discovered that the real costs are a fraction of what the carriers claimed; this is really changing our whole thinking about networking."

The convergence of metropolitan fibre networks from the likes of MCI Worldcom, AAPT, Primus, and PowerTel, with long-haul intercapital networks from companies such as FiberTel/Amcom, will soon bring similar benefits to Australia. And as increasing competition saturates the country with fibre-optic backbones, local companies will gain the ability to link branch offices in both city and country areas using high-speed WAN connections rated in hundreds of megabits per second.

Once WANs match the speed of internal network backbones, companies will totally reconsider networking strategies long predicated on the limitations of low-speed inter-office links. A coming breed of routers with built-in optical networking interfaces will be able to dynamically request additional 155 Mbps STM-1 trunks, for example, to meet periods of peak demand. Even more importantly, such high speeds mean that functions such as remote backup and centralised enterprise application hosting can be provided to any branch office with absolutely no speed degradation.

Pundits predict this fact will lead many companies to strip remote offices of complex and unnecessary IT gear. In a few years, a typical office might instead consist of myriad desktop computers linked to simple workgroup switches, which in turn connect directly onto the fibre and back to a single data centre at the head office. Employees using those desktops will access corporate data and applications directly across the WAN at speeds that would today make their heads spin. Secure interconnects between business partners will provide seamless links that truly bridge the gaps between business partners by enabling LAN-speed collaboration among dynamic virtual workgroups.

Much as Cabletron led the industry by launching pre-standard Gigabit Ethernet equipment back in 1997, Enterasys plans to ship switches supporting up to seven pre-standard 10 GE optical interfaces by the end of this year. Confluence on 10 GE, which retains none of Ethernet’s design except for its packet-based transmission, could eventually sideline previous high-speed WAN contender ATM (Asynchronous Transfer Mode).

ATM equipment is still quite expensive after nearly a decade of development, since technical excellence hasn’t resulted in the market domination that was once predicted. For most companies, its ability to provide high-granularity quality of service will be overshadowed by its relative complexity and the tremendous cost efficiencies of 10 GE. ATM will still be supported in particular segments over direct SONET/SDH cores, but the majority of companies will adopt IP-based services provided over DWDM-based fibre cores using 10 GE interconnects.

"For once in the history of this industry, WAN bandwidth is outstripping the capacity requirements of the LAN," says John Roese, US-based chief technology officer of Enterasys, Cabletron Systems’ recently created enterprise networking arm.

"We see this concept of infinite WAN bandwidth as being far more than same model we use today but faster. It’s going to drive new models of where companies put resources in relation to their customers and where they’re using them. We already have large customers evaluating the concept of server-less buildings, which they’ve never been able to do with a traditional TDM WAN service. Now, the fact that they can put 10 Gbps or more across their WAN changes the paradigm. At the end of the day, the vision is to be able to have infinite optical bandwidth available when and where you need it."

Back to Best Technical 
Top of page

Content Copyright © the author/publisher listed above

Design Copyright © Consensus Pty Ltd

This web-site uses frames, click here for the full picture