Gridless

Part 3: ROADM and control plane developments

ROADMs and control plane technology look set to finally deliver reconfigurable optical networks but challenges remain.

Operators are assessing how best to architect their networks - from the router to the optical layer - to boost efficiencies and reduce costs.  It is developments at the photonic layer that promise to make the most telling contribution to lowering the cost of transport, a necessity given how the revenue-per-bit that carriers receive continues to dwindle.

Global ROADM forecast 2009 -14 in US $ miliions Source: Ovum 

“The challenge of most service providers largely hasn’t changed for some time: dealing with growth in demand economically,” says Drew Perkins, CTO of Infinera. “How can operators grow the capacity on each route and switch it, largely on a packet-by-packet basis, without increasing the numbers of dollars going into the network.”

Until now, dynamic optical networking has been conducted at the electrical layer. Electrical switches set up connections within a second, support shared mesh restoration in under 100 milliseconds (ms), and have a proven control plane that oversees networks up to 1,000 networking nodes.  This is the baseline performance that a photonic layer scheme will be compared to, says Joe Berthold, vice president of network architecture at Ciena.

AT&T’s Optical Mesh is one such electrically-enable service. Using Ciena’s CoreDirector electrical switches, customers can change their access circuits in SONET STS-1 (50 Megabits-per-second) increments via a web interface. AT&T wants to expand further the capacity increments it places in customers’ hands.

“The real problem with operators today is that it takes way too long to set up a new connection with existing optical infrastructure.”

Tom McDermott, Fujitsu Network Communications

Developments at the photonic layer such as advances in reconfigurable optical add-drop multiplexers (ROADMs) as well as control plane and management software complement the electrical layer control. ROADMs enable the redirection of large bandwidths while the electrical layer, with its sub-wavelength grroming and switching at the packet level, accommodates more rapid traffic changes. Operators will benefit as the two layers are used more efficiently. 

“The photonic layer is the cheapest per bit per function, cheaper than the transport layer – OTN (Optical Transport Network) or the SONET layer ­– and the packet layer,” says Brandon Collings, CTO of JDS Uniphase’s consumer and commercial optical products division. “The more efficient, functional and powerful the control plane, the better off operators will be.”

ROADM evolution

ROADMs sit at the core of the network and largely define its properties. “The network’s wavelengths may be 10, 40 or 100 Gig; that is just bolting something on the edge. The ROADM sits in the middle, it’s there, and it has to handle whatever you throw at it,” says Simon Poole, director new business ventures at Finisar.

Operators have gone from using fixed optical add-drop multiplexers (OADMs) to ROADMs with fixed add-drop ports to now colourless and directionless ROADMs.  Each step increases the flexibility of the switching devices while reducing the manual intervention required when setting up new lightpaths.

“There has been a much greater drive in the US [for ROADMs] but it is now picking up in Europe”

Ulf Persson, Transmode Systems

Network architectures also reflect these advances. First ROADMs were typically two-dimensional nodes that enabled metro rings. Now optical mesh networks are possible using the ROADMs’ greater number of interfaces, or degrees.

Transmode Systems has many of its customers – smaller tier one and tier two operators - in Europe, and focusses on the access, metro and metro-regional markets.

“It is not just the type and size of the operators, there are also regional differences in how all-optical and ROADMs are used,” says Ulf Persson, director of network architecture at Transmode. “There has been a much greater drive in the US [for ROADMs] but it is now picking up in Europe.” One reason for limited ROADM demand in Europe, says Persson, is that for smaller networks it is easier to design and predict growth.

Operators with fixed OADMs must plan their networks carefully. When provisioning a service, engineers have to visit and reconfigure the nodes needed to support the new route. In contrast, with fixed add-drop port ROADMs, engineers only need visit the end points.

The end points require manual intervention since the ROADM restricts the lightpath’s direction and wavelength. ROADMs at nodes along the route can at least change the direction but not the lightpath’s wavelength. “You can do the express routing efficiently, it is the [ROADM] drop side that that is not automated at this point,” says Poole.

This still benefits operators even if it doesn’t meet all their optical layer requirements. “Where ROADMs have helped is that while service technicians must visit the end points - connecting the transponder card and client equipment -  they save on the intermediate site visits,” says Jörg-Peter Elbers, vice president, advanced technology at ADVA Optical Networking. “Just by a mouse click, you can set up all the nodes in the right configurations without the hassle of doing this manually.”

The result is largely static networks that once set up are seldom changed. “The real problem with operators today is that it takes way too long to set up a new connection with existing optical infrastructure,” says Tom McDermott, distinguished strategic planner, Fujitsu Network Communications.

Colourless and directionless ROADMs aim to solve this. A tunable transponder can now be pointed to any of the ROADM’s network interfaces while exploiting its tunability for the lightpath’s wavelength or colour.

The ROADM percentage of the total metro WDM market. The market comprises Coarse WDM, fixed and reconfigurable OADMs. Source: Ovum

Such colourless and directionless ROADMs offer several benefits. An operator can have several transponders ready for provisioning to meet new service demand. This arrangement for ‘deployment velocity’ has yet to take hold since operators are reluctant to have costly transponders idle, especially if they are 40 and 100 Gigabit-per-second (Gbps) ones.

Colourless and directionless ROADMs will more likely be used for network restoration during maintenance or a fibre cut. This is slow restoration, nowhere near the 100ms associated with electrical signalling; optical signals are analogue and each lightpath must be turned up carefully. “When you re-route an optical signal going more than 1,000km, taking it off one route affects all the other signals on that route and they need to be rebalanced; then putting it on another route, they too need to be rebalanced,” says Infinera’s Perkins. “It is very difficult to manage.”

ADVA Optical Networks cites the example of operators using 1+1 route protection. When one route is down for maintenance, the remaining route is left unprotected. Colourless and directionless ROADMs can be used to set up a spare route during the maintenance.  In developing countries, where fibre cuts are more common, 1+N protection can provide operators with redundancy to survive multiple failures. Such a restoration strategy is especially needed if getting to a fault in remote area may take days.

The extra flexibility of the newer ROADMs provide will also aid operators with load balancing, moving traffic away from hotspots as the network grows.

WSSs at the core

The building block at the core of the ROADM is the wavelength-selective switch (WSS). Such switch building-blocks are implemented using light-switching technologies such as liquid crystal, liquid crystal on silicon and MEMS. The WSS routes a lightpath to a particular fibre, with the WSS’s degree in several configurations: 1x2, 1x4, 1x9 and, under development, a 1x23. The ROADM’s degree relates to the network interfaces it supports - a 2-degree ROADM supports two fibre pairs, pointing east and west. A WSS is used for each ROADM degree and hence with each fibre pair. 

A 1:9 WSS supports an 8-degree ROADM with the remaining two ports used for local multiplexing and demultiplexing.  So far, eight fibre pairs have been sufficient. A 1:23 WSS is being developed to support yet more degrees at a node. For example, more than one fibre pair can be sent in the same direction (doubling a route’s capacity) and for adding extra add-drop banks. JDS Uniphase is one vendor developing a 1:23 WSS.

"Where I’m seeing first interest beyond 1x9 is business service or edge applications - where at a given node point operators need a lot of attachments to different enterprise networks at high capacity multi-wavelength levels,” says  Tom Rarick, principal engineer, transport, at Tellabs. “Within infrastructure applications, a 1x9 provides the degree of fibre connectivity necessary; I rarely see beyond a 1x6.”

“The next big thing [after colourless and directionless] is what people call contentionless and gridless, adding yet more flexibility to the optical infrastructure.”

Jörg-Peter Elbers, ADVA Optical Networking 

For a WSS, the common port connects to the outbound fibre of any given direction, whereas the 9 ports [for a 1:9] face inwards to the centre of the node, says JDSU’s Collings (click here for a JDSU presentation).  He describes the WSS as a gatekeeper that determine which lightpaths from which fibres leave a node. As for the incoming fibre, the lightpaths carried are sent to each of the other ROADM node WSSs, one for each direction. “Each WSS selects which channel from which port leaves the node,” says Collings.

To make a ROADM colourless and directionless, extra add-drop ports hardware must be added. These route lightpaths to any node and on any wavelength, or drop any lightpath from any of the other nodes. The add-drop node is built using further WSSs.  

One issue still to be resolved is whether WSS sub-system vendors provide all the elements that are added to the WSS to make the ROADM colourless and directionless. “It is not clear whether we should be developing the whole thing or providing the modules for the customers’ line cards and systems,” says Poole.

“The next big thing [after colourless and directionless] is what people call contentionless and gridless, adding yet more flexibility to the optical infrastructure,” says Elbers.(Click here for an ADVA Optical Networking ROADM presentation)

Contentionless refers to avoiding same-wavelength contention. With a colourless and directional ROADM, only one copy of a particular wavelength can be dropped. “With a four-degree node you can have a 96-channel fibre on each,” says Krishna Bala, executive vice president, WSS division at Oclaro. “You may want to drop at the node lambda₁ from the east and the same lambda₁ coming from the west.” To drop the two, same wavelengths, an extra add-drop block is required.

To make the ROADM fully contentionless, as many add-drop blocks as ROADM degrees are needed. This requires more WSSs, or alternatively 1:N splitters, as well as N:1 selection switches. This way any of the dropped wavelengths from any of the incoming fibres can be routed to any one of the add-drop’s transponders.

“The building blocks for colourless and directionless ROADMs are there; we sell them as a product,” says Elbers, who stresses that the cost of the WSS building block is coming down. But the question remains whether an operator values such network functionality sufficiently to pay.  “Without naming names, big carriers are looking at these – they want to have a future-proof, simple-to-plan network,” says Elders.

“It is strictly economics,” agrees Ciena’s Berthold. “We’ve offered a colourless-directionless ROADM for some time. Some buy that but more often they are going for lower cost.”

Gridless

A further ROADM attribute being added to the WSS is gridless even though it will be several years before it is needed. WSS vendors are keen to add the feature – adaptive channel widths - now so that operators’ ROADM deployments will be future-proofed.

“They [WSS vendors] have got to be in a tough spot. They have to invest all that [R&D] money while they [carriers/ system vendors] ask for the world.”

Ron Kline, Ovum.

Channel bandwidths wider than 50GHz will be needed for line speeds above 100Gbps. Gridless refers to the ability to accommodate lightpaths that do not just fit on the International Telecommunication Union’s (ITU) 50 or 100GHz grid. WSS makers are developing fine pass-band filters that when combined in integer increments form variable channel widths.

“There is a great deal of concern from operators about how they can efficiently use the spectrum to maximize fibre capacity,” says Poole. What operators want is the ability to generate channel bandwidth with much finer granularity and to move away from fixed channel widths.

According to Poole, NTT have demonstrated a ROADM with 12.5GHz increments, others are thinking 25GHz or even 37.5GHz. Finisar says this issue has gained much operator attention in the last six months and that there is urgency for WSS vendors to implement gridless so that any ROADM deployed will be able to support future transmission rates beyond 100Gbps.

“They [WSS vendors] have got to be in a tough spot,” says Ron Kline, principal analyst for network infrastructure at Ovum. “They have to invest all that [R&D] money while they [carriers/ system vendors] ask for the world.”

Coherent receiver technology used for 100Gbps optical transmission will also help enable dynamic optical networking by overcoming technical issues when rerouting paths.

Optical signal distortion in the form of chromatic dispersion and polarisation mode dispersion (PMD) are so much worse at 40Gbps and 100Gbps. Even on 10Gbps routes, where tolerance to dispersion is greater, compensation can be an issue when redirecting a lightpath during network restoration. That is because the alternative route is likely to be longer. Unless the dispersion compensation is correct, there is uncertainty as to whether the alternative link will work, says Ciena’s Berthold.

“With a coherent receiver, you are now independent of dispersion since you can adaptively compensate for dispersion using the [receiver’s] DSP ASIC,” says Berthold. “You no longer have to worry is you have it [the compensation tuning] just right.”

The ASIC can also deliver real-time latency, chromatic dispersion and PMD network measurements at path set-up. This avoids first testing the link, and possible errors when entering measurements in the planning-path network set-up tools. “Coherent technology for 40 Gig and 100 Gig is potentially a game changer in making ROADMs work,” says Berthold.

Coherent digital transponders at 40 and 100Gbps will also drive the deployment of more advanced ROADMs, argues Oclaro. “The need to extract value from the bank of [40 and 100G coherent] transponders in a colourless-directionless sense becomes a lot more important,” says Peter Wigley, director, marketing and technology strategy at Oclaro.

Control plane

Tunable lasers, flexible ROADMS and even coherent technology may be prerequisites for agile optical networks, but another key component is the control plane software. “Many of the networks today have some of the hardware components to make them agile but lack the software,” says Andrew Schmitt, directing analyst, optical, at Infonetics Research. (See ROADM Q&A with Andrew Schmitt.)

The network can be split into the data plane, used to transport traffic, the control plane that uses routing and signalling protocols to set up connections between nodes, and the management plane that oversees the control plane.

“What is deployed mostly today is a SONET/SDH control plane,” says Tellabs’ Rarick. “This is to manage SONET/SDH ring or mesh networks, using standalone cross-connects or partnered with ROADMs, with the switching primarily done electrically.”    

Three industry bodies are advancing control plane technology in several areas including the optical level.

The Internet Engineering Task Force (IEFT) is standardising Generalized Multiprotocol Label Switching (GMPLS) while the ITU is developing control plane requirements and architecture dubbed Automatically Switched Optical Networks (ASON). The third body, the Optical Internetworking Forum (OIF) oversees the implementation efforts.

“The [GMPLS/ASON] control plane comprises a common part and technology-specific part,” says Hans-Martin Foisel, OIF president. The specific technologies include SONET/SDH, OTN, MPLS Transport Profile (MPLS-TP) and the all-optical layer.

 "The more efficient, functional and powerful, the control plane, the better off operators will be"

Brandon Collings, JDS Uniphase.

“Using a control plane with all-optical is a challenge,” says Foisel. “The control plane has to have a very simplified knowledge of the optical parameters.” The photonic layer has numerous optical parameters that can be used. Any protocol needs to streamline the process such that simple rules are used by operators to decide whether a route can be completed or whether signal regeneration is needed.

The IETF is working on wavelength switched optical networks (WSON), the all-optical component of GMPLS, to enable such simplified rules within a single network domain. “GMPLS cannot control wavelengths today using ROADMs and that is what is being standardised in WSON,” says Persson.

What is beyond of scope of WSON is routing transparently between vendors, says Foisel.  "It is almost impossible to indentify all the optical parameters in an inter-vendor way for operators to fully use,” says Fujitsu’s McDermott. "You end up with a huge parameter set."

So what will the photonic control plane look like?

“The whole architecture of control will be different that what is done in the electrical domain,” says Berthold. It will combine three main functions. One is embedded intelligence that will learn fibre-route characteristics and optical parameters from the network, data which will use be by each vendor in a proprietary way. Another is a propagation modelling planning tool that will process data offline to determine the viable network paths. These paths will then be preloaded into network elements as well as recommendations as to the preferred ones to use to avoid contention. Finally, use will be made of the signalling to turn these paths up as rapidly as possible. “This is certainly not the same model as electrical,” says Berthold.

By combining electrical and optical switching, operators will be able to continually optimise their networks. “They can devolve their networks to the lowest cost and most power-efficient solution,” says Berthold.

Ciena, for example, is adding colourless-directionless ROADMs to its 3.6 terabit-per-second 5430 electrical switch. “When you start growing traffic from a low level you need electrical switches in many places in order to efficiently fill wavelengths,” says Berthold. “But as traffic grows there is more opportunity to bypass intermediate nodes with an optical path.” By tying the ROADM with the electric switch, traffic can be regroomed and electrical paths set up on-demand to continually optimise the network.

Challenges

Despite progress in ROADM hardware and control plane management, challenges remain before a remotely controlled all-ROADM mesh network will be achieved.

One is handling customer application rates at 1, 10 and 40Gbps on 100 Gbps infrastructure. This will use the OTN protocol and will require electrical switch and control plane support.

Interoperability between vendors’ equipment must also be demonstrated. “Interlayer management – it is not enough just to do optical,” says Kline. “And it is not only between layers but interoperability between vendors’ equipment.” Thus, even if Verizon Business is correct that colourless, directionless ROADMs will become generally available in 2012; the vision of a dynamic optical network will take longer.

“Coherent technology for 40 Gig and 100 Gig is potentially a game changer in making ROADMs work”

Joe Berthold, Ciena

“GMPLS/ASON are still years out and some operators may never deploy them,” says Schmitt at Infonetics. But Kline highlights the vendors Huawei and Alcatel-Lucent as keen promoters of control-plane-enabled dynamic optical networking.

“Huawei has 250 ASON applications with over 80 carriers, and 30-plus OTN WDM ASON applications,” says Kline. Here, an ASON application is described a ring or nodes that use a control plane for automated networking.  “These are small and self-contained; not AT&T’s and Verizon’s [sized] meshed networks,” says Kline, who adds that Alcatel-Lucent also has such ASON deployments.

There are also business-case hurdles associated with photonic switching to be overcome.

Doing things on-demand may be compelling but need to be proven, says Jim King, executive director of new technology product development and engineering at AT&T Labs. This is easier to prove deeper in the network. “In the middle of the network it is easy because the law of large numbers means I know I need lots of capacity out of Chicago, say; I just don’t know whether it needs to go north, east or south,” says King. “But when a just-in-time delivery requirement extends to the end of the network, the financials are much more challenging based on how close you need to get to customer premises, cell towers or critical customer data centres.”

What next?

Oclaro too believes that it will be another two years before colourless, directionless and contentionless ROADMs start to be deployed in volume. The challenge thereafter is driving down their cost.

Another development that is likely to emerge after gridless is faster switching speeds to reduce network latency. Operators are using their mesh networks for restoration but there is a growing interest in protection and restoration at the optical layer, says Bala. “We were seeing RFPs (request-for-proposals) where a WSS of below 2s was ok,”’ he says. “Now it is: ‘How fast can you switch?’ and “Can you switch below 100ms’.”

This is driving interest in optical channel monitoring. “Ultimately it will require the ability to monitor the ROADM ports from signal power and to detect contention, and you’ll need to do this quickly,” says Wigley. “It is not very useful having a fast WSS unless you know quickly where the traffic is going.”

Technology will continue to provide incremental enhancements. The cost-per-bit-per-kilometer has come down six or seven orders of magnitude in the last two decade, says Finisar’s Poole. Apart from the erbium-doped fibre amplifier (EDFA), no single technology has made such a sizable contribution. Rather it has been a sequence of multiple incremental optimisations. “Coherent technology is one way of getting more data down a pipe; gridless is another to get a 2x improvement down a pipe,” says Poole. “Each of these is incremental, but you have to keep doing these steps to drive the cost-per-bit down.”

Meanwhile operators will look to further efficiencies to keep driving down transport costs. “Operators are looking at tradeoffs of router versus optical switching,” says McDermott. “They are going through various tradeoffs, the new services they might offer, and what is a flexible but cost effective solution.” As yet there is no universal agreement, he says.

“The balance between the two [the optical and electrical layer] is the key,” says Infinera’s Perkins. “There is a balance you have to reach to achieve the best economics: the lowest cost network supporting the highest capacity possible at a cost you can afford, and operate it with the fewest people.”

And ROADMs will be deployed more widely. “In 3-5 years’ time everything will have a ROADM in it – it better have a ROADM in it,” says Kline. At the electrical layer it will be Ethernet and at the optical it will be OTN and lightpaths. “It is all about simplification and saving costs.”

Other dynamic optical network briefing sections

Part 1: Still some way to go

Part 2: ROADMS: reconfigurable but still not agile