Still some way to go

Briefing: Dynamic optical networking 

Part 1: The vision .... back in 2000

I came across this article (below) on the intelligent all-optical network. I wrote it in 2000 while working at the EMAP magazine, Communications Week International, later to become Total Telecom.

What is striking is just how much of the vision of a dynamic photonic layer is still to be realised.  Back then it had also been discussed for over a decade. And bandwidth management, like in 2000, is still largely at the electrical layer.

And yet much progress has been made in networking technology. But the way the network has evolved means that a more flexible photonic layer, while wanted by operators, is only one aspect of the network optimisation they seek to reduce the cost of transporting bits.

The second and third parts of the dynamic optical networking briefing will discuss how often operators reconfigure their networks and what is required, as well as developments in reconfigurable optical add-drop multiplexer (ROADM) and control plane technologies that promise to increase the flexibility of the photonic layer.

  --+++--

Seeing the light (April 17th, 2000)

The next generation of networks is coming, with abundant bandwidth and flexible services available on-demand, and intelligent management and provisioning at the optical layer. Roy Rubenstein finds out what's in store and who's set fair in this optical future.

For all its air of novelty, all-optical networking is actually a mature idea. Discussed for the best part of a decade, all-optical networks have perennially promised to deliver the next generation of “intelligent” services, yet besides the stir caused by the arrival of dense wavelength division multiplexing (DWDM), forcing greater capacity over fiber networks, there has been little in the way of tangible development.

Now, with limited ceremony, optical networking is reasserting itself, and the signs are that you could reap the benefits sooner than you think.

What excites operators most is the prospect of bandwidth on demand: high-speed links set up with little more than a few mouse clicks. But the technology is creating dilemmas as well as opportunities. On the one hand, the newer operators can enter the market with a sleeker network - fewer layers and fewer nodes - accompanied by the latest billing and management software. On the other hand, incumbent operators are facing the dilemma of when to embrace the technology and how to integrate it with their legacy equipment.

“Most of the network planners agree this is the way to go,” says Barry Flanigan, senior consultant at Ovum Ltd., of London. "The question is the precise technology and timing."

Flexible bandwidth

Flexible bandwidth provisioning will enable a range of services that have not been practicable until now. For example, network planners in corporations will no longer have to guess - and live with the consequences - each time they budget their capacity requirements and agree horribly rigid contract terms.

In fact all manner of on-tap services become possible when bandwidth is set up and collapsed on an hourly or minute-by-minute basis. One example is bandwidth trading between carriers, enabling operators with their own networks to grab business such as voice services while demand is there, and off load capacity when it is not.

A further example is the broadcasting of sporting events. Instead of satellite coverage, a TV company could set up a cheaper terrestrial network link to each sporting venue, but provision capacity only for the duration of the event. And content providers can offer services locally. Opening pipes, a provider can download and store video on demand on a country-by-country basis ready for delivery, before closing the links.

“That way the service seems a lot quicker,” says Andy Wood, chief technology officer at Storm Telecommunications Ltd., based in London.

Adding intelligence

The key to this flexible bandwidth provisioning is optical switches, which introduce “intelligence” to the optical layer. An optical switch-whether electrically based or all-optical-routes complete wavelengths of light packed with up to 10 megabits of data.

“The scenario today is that bandwidth management is at the electrical layer,” says Richard Dade, director for industry liaison, optical networking group, at Lucent Technologies Inc., Murray Hill, New Jersey. “By the end of this year-2001 it will transition into the optical layer.”

This is also the view of Nick Critchell, product marketing manager for core optical internetworking products at San Jose, California-based Cisco Systems Inc. “Looking forward two years to the core routing, it will provide intelligent switching and intelligent restoration,” he says.

But others question the impact such technologies are having on the awareness of the underlying optical network. “Intelligence may be too strong a word for it,” says Dr. David Huber, chief executive of Corvis Corp., the Columbia, Maryland-based optical networking technology start-up.

What interests him is the sheer data traffic-handling capabilities--transporting terabits of data--and network efficiencies that all-optical switches promise. For example Huber predicts network utilization will exceed 80% using all optical switches. Current network utilization figures are below 50%.

When it comes to the operators, it seems the newer breed is keenest to embrace the technology. For them, adopting intelligent optical switching provides a simpler network, removing the need for Sonet/Synchronous Digital Hierarchy (SDH) transmission equipment. They also gain in reduced operating costs and system efficiencies through the use of the latest network operating system, billing and management systems.

Established operators, in contrast, have an enormous legacy of network equipment. “Different telecoms operators have different levels of awareness [in adopting intelligent optical networks],” is the view of Margaret Hopkins, principal analyst at Cambridge, England-based consultancy Analysys Ltd.

And Hopkins is quick to stress that whatever the merits of the latest optical switching, it will not cause more established technologies to disappear any time soon. “Sonet gives you very fast reconfiguration [if a fiber is cut],” says Hopkins, pointing out that optical networks have some way to go before assuming this role. “For other users, SDH performs functions such as mixing different types of traffic - pulse coded modulated voice and IP - on the one wavelength. This is important, because a wavelength is an awful lot of capacity."

Some way to go

The future of Sonet/SDH is also secure while voice-over-Internet protocol (VoIP) traffic remains low, particularly as a proportion of all voice traffic. “With voice on packet networks, growth has been modest from a European perspective,” says Eric Owen, London-based senior director for European telecommunications at International Data Corp., Framingham, Massachusetts. “When asked about VoIP - medium-to-large enterprises across Europe - only 4% to 5% are doing it,” he says.

But operators are aware that they cannot afford to ignore intelligent optical switches, and several are already trialing the technology including MCI WorldCom Inc. and Williams Communications Inc.

One next-generation carrier has been bolder still. “The first indication of an optical switching network is the announcement from Storm,” says Chris Lewis, managing director of research and consulting at the Yankee Group Europe, of Watford, England. “Storm is basing its case on being able to switch in bandwidth pretty quickly,” he adds.

Storm, a carriers’ carrier, announced last month that it has acquired $100 million-worth of dark fiber to which it will connect optical equipment from Chelmsford, Massachusetts-based Sycamore Networks Inc. “It's the first example we've come across of concrete plans,” says Lewis.

The significance of Storm's announcement is the promise of bandwidth on demand. Mark Stewart, Storm Telecom's business development director, says its network users - carriers, large corporations and Internet service providers - can have the bandwidth they require, with costs based on usage. “A customer may need five STM-1, 155-megabit-persecond links one week and nine the next,” says Stewart.

Currently, an STM-16 (2.5 gigabits per second) link must be leased long-term to guarantee capacity, but through Storm users can buy the bandwidth they need in increments as small as 45 megabits per second, available for lease for “a short period,” according to the company. Storm has yet to finalize its service details, but claims 75% of its network will be up and running by the end of the summer.

The budding pan-European carrier is an example of what Analysys’ Hopkins refers to as a newer operator “configuring more interesting services more quickly.”

For “older” operators, meanwhile, expanding capacity involves adding overlays to their networks. Investing in the latest networking technology is seen as a strategic move, which needs to be taken, but which cannot be implemented in one fell swoop. “They have a share price - they don’t want to be seen to be a dinosaur,” says Hopkins.

And when they do add the latest IP technologies to their infrastructure, they “don’t get the simplification benefits,” she says.

Key stage

Equipment vendors, meanwhile, share the view that flexible switching is a key stage in the evolution of an all-optical layer.  “All vendors are working on this: not just to get capacity but to exploit these wavelengths,” says Ovum's Flanigan.

At present the bulk of the public network is still based on Sonet/SDH transmission technology, to which DWDM has been added to meet the demand for IP traffic. “[U.S. long-distance operator] Sprint now has 80% of their routes on WDM,” says Bill Anderson, director of optical networking research at Morristown, New Jersey-based Telcordia Technologies Inc.

Yet while it has fulfilled a need in addressing the steep demand for capacity in recent years, DWDM is some way from being the solution to flexible, “intelligent” bandwidth provisioning. “DWDM can be seen as large, relatively dumb pipes,” says Anderson.

The issue, according to Rick Dodd, director of marketing strategy at optical networking specialist Ciena Corp., of Linthicum, Maryland, stems from the nature of Sonet/SDH technology.

“A lot of human intervention has been traded for intelligence,” he says, referring to the economic realities when SDH technology was first introduced around 10 years ago. Then, the manual setting up of links made sense, but with a decade of improvement in the performance/cost of microprocessors and memory, this is no longer the case.

Economical touch

“It’s very economical to add software and application-specific integrated circuits to the optical infrastructure to deliver a whole new type of network,” says Dodd.

Sycamore Networks is one vendor seeking to exploit this shift. It describes itself as a developer of hardware and software for next-generation optical networks. “What we call intelligent optical networks,” says Jeff Kiel, Sycamore's vice president of product marketing.

Kiel points out that currently Sonet/SDH equipment converts the data into optical form and manages traffic across the network. Sycamore Networks, for one, argues that with intelligent switch technology operators no longer even have to buy SDH equipment.

“You can provide the SDH framing - the same functionality - but subsume it into the optical layer,” says Kiel. Instead of having an IP device connecting to SDH equipment, which in turn connects to DWDM and ultimately fiber, the number of layers in the network can now be limited to three: the IP device linking directly to the intelligent optical network, which in turn connects to the fiber.

Lucent's Dade agrees that SDH can be bypassed, but believes, like Analysys’ Hopkins, that its course has not been run just yet.

“Our projections are that there is still a long life for Sonet/SDH services,” he says. “There will be strong growth for Sonet/SDH and stronger growth for intelligent optical network equipment.”

Carriers and manufacturers alike are looking to collapse telecoms networks into two layers. At the upper, switched layer, asynchronous transfer mode (ATM) and IP are being consolidated and are pulling in certain core SDH functions. The challenge is to ensure that the reconfiguration at the optical layer matches the tried-and-tested reliability of the SDH layer.

Having intelligent switching devices at the optical level is a precondition of this, and raises the prospect of a network with the intelligence to adapt itself--what Dennis Jennings, Telcordia Technologies’ vice president for next-generation networks, calls a dynamic reconfigurable network.

The Optical Domain Service Interconnect (ODSI) initiative [the ODSI merged with the OIF’s signaling workgroup], set up in January [2000] by 50 vendors and service providers, is working to develop a signaling scheme to ensure better communication between the two layers. “In the next year or two carriers will replace networks with two layers,” predicts Jennings.

One consequence of the increasing awareness at the optical layer is that assigning bandwidth may no longer be the preserve of operators. If the ODSI initiative proves successful, an IP router will be able to request an extra wavelength from the optical layer whenever it detects congestion, and relinquish it once the data surge passes.

“Optical networking will enable the provisioning of bandwidth to be instantaneously available,” says Alan Taylor, consulting engineer for Europe at backbone router company Juniper Networks Inc., of Mountain View, California.

“The ODSI is focused on a very simple problem,” says Rick Thompson, senior product marketing manager at Sycamore Networks, one of the co-founders of the initiative. “[The problem] has been deliberately scaled down with the aim of a quick deployment.”

Current routers can view the bandwidth between points, but no signaling scheme is available to enable an extra wavelength to be requested. ODSI intends to extend the signaling of the Multiprotocol Label Switching protocol (MPLS), which is used by routers to establish a path through the network for a given packet stream.

“This signaling between the data and the optical layers promises dynamic self-reconfiguring networks rather than ones needing human intervention,” claims Telcordia's Jennings.

A framework document has been completed and was discussed at an ODSI meeting held in Chicago on 6 April. Completion of the signaling extension is expected within the next three months.

ODSI is not developing a standard, but rather an open framework which the initiative’s members will adopt. Interoperability testing will then ensue.

“You can expect large scale deployment in 2001,” says Thompson.

The equipment vendors

The latest optical switch from Sycamore Networks Inc., the SN 16000, illustrates how optical nodes are gaining in awareness. Switching wavelengths - a necessary requirement if bandwidth is to be provisioned ‘in Seconds’ - is just one of its attributes. The node can also discover a network’s topology, so that should a fiber linking two cities be cut, the traffic can be rerouted on a different path.

This “intelligence” can even take into account the dispersion characteristics of the fiber. Thus if the original path is based on a Sonet OC-192 link, the switch redistributes the data accounting for the fiber’s nature.

Williams Communications Inc., a carrier based in Tulsa, Oklahoma, is trialing an intelligent optical switch from Ciena Corp. in addition to adopting Sycamore’s equipment. Dubbed Multiwave Coredirector, Ciena’s switch is also network-aware, and through the use of the optical signaling and routing protocol (OSRP) can communicate with other Ciena optical switches to determine the network’s status as well as planning the best path for traffic.

“We are enabling optical carriers to deliver optical bandwidth very quickly. In a variety of sizes--wavelengths and fractions of wavelengths - and employed with a variety of priorities,” says Ciena's Dodd. According to the vendors this flexibility can boost carriers’ revenues and have an impact on the implementation cost. “If you look at where service providers are spending their money, for every $1 spent on equipment, $2 to $4 is spent making it run,” says Dodd.

Williams’ appetite for trialing technology is not confined to Ciena and Sycamore's electrical-based optical switches. It is also trialing an all-optical cross-connect switch from Corvis Corp.

According to the carrier not only will an all-optical switch ease the huge increase in expected traffic, it will also help meet the variable demand in bandwidth as traffic becomes increasingly unpredictable, with applications such as video broadcast across the Internet becoming commonplace.

Andy Wright, chief technologist for optical networking at Williams describes Corvis’ technology as suited to “ultra long-haul networking,” removing the need for regenerators as the wavelengths are routed from one end of the optical layer to the other.

Regeneration equipment for a channel can cost between $60,000 and $70,000, says Wright. For a typical 100-channel site the cost is $7 million. By eliminating regeneration, what is required is a $100,000 optical amplifier, “a scale change in prices.”

Qwest Communications Corp., meanwhile, believes all-optical networking will drive down its operating costs by as much as 70%.

Lucent Technologies Inc. is another vendor developing an all-optical switch. “In our view the optical-electrical-optical conversion when routing a wavelength should not happen,” says Dade. Having an all-optical implementation is not only intrinsically cheaper, he says, but extraordinarily flexible. Lucent's Lambdarouter switch is currently being trialled by a U.S. long-distance operator and will be in production by year end.

But the view of Cisco Systems' Nick Critchell is that all-optical switching has still to prove it is reliable enough to be deployed across optical networks. That said, he is in no doubt about the technology’s merit. “We are moving [to develop an all-optical switch] and will be in that space,” he says.

Other sections of this briefing:

Part 2: ROADMS: reconfigurable but still not agile

Part 3: To efficiency and beyond