Controlling the light with next generation ROADM architecture

ROADM technology offers service providers an efficient way to scale current metro-regional infrastructures and paves the way for new revenue opportunities. Deployment is now just a question of when

During the manic years between 1999 and 2000, service providers—encouraged both by projections of exponential demand growth and equipment providers who promised performance that defied the laws of physics and economics—looked at building dynamic optical networks that would enable them to not only deliver wavelength services but to further enhance profitability by offering time-based sharing of the transmission infrastructure.

These network architectures delivered flexibility and were reconfigurable because they relied on a relatively static transmission infrastructure of either simple fiber, point-to-point dense wavelength division multiplexing (DWDM) or fixed optical add/drop multiplexing (OADM), with network reconfiguration being accomplished by large all-optical switches that could be located strategically in the network.

Since then, service providers have struggled to make a viable business case for the ability to reconfigure at the optical layer, basing the business case primarily on the expectation of new revenues from wavelength services―revenues that have yet to materialize. As service providers struggle to meet basic network growth demands at reasonable price points, moving to a reconfigurable OADM optical layer (ROADM) in metro-regional networks is an attractive solution with staying power, offering both lower costs for basic transport, as well as paving the way to new revenue opportunities.

Doug Green, principal at The Bradam Group, an optical telecom consultant group based in Gainesville, VA, suggests that early ROADM technologies introduced as many problems as they solved. The basic problem was in the rebalancing and tweaking of network elements following a change to the routing of wavelengths in the network. Since service providers were unable to tweak what couldn’t be measured without demultiplexing the entire data stream, the technology became inapplicable.

Today’s ROADM systems introduce the ability to deploy and operate reconfigurable wavelength transmission technology without introducing these problems. In fact, Green says, ROADM is the most promising telecommunications technology to come along in years and appears to be the most viable new architecture for future optical builds.

Evolution to ROADM

The traditional Sonet/SDH multi-node ring transmission infrastructures have served the metro transport market extremely well and with good reason. With trunk capacities being dramatically larger than the tributary sizes required to be added and dropped at each location, the Sonet ring provides an efficient mechanism to aggregate, protect and transport traffic with significant flexibility in where the bandwidth can be routed. With the high trunk/tributary bandwidth ratios of voice services dominating the demand profile, these ring-based transport networks could flexibly and cost-effectively deal with all the demand growth, churn and reconfiguration that typify the somewhat chaotic metro transport market.

Over the last three years, ROADM has transitioned from being too expensive for a metro environment to having a compelling business case that is turning the heads of many metro-regional service providers.

But as data services continue to drive point-to-point bandwidth demands to 2.5 Gb/s and 10 Gb/s speeds, this infrastructure becomes inefficient, expensive and eventually leads to fiber exhaust. The fundamental problem here is that trunk speeds have not grown as fast as tributary demand, and there is little hope that cost-effective 40 Gb/s Sonet ADMs will save the day.

Even if we had them now, moving to a 40 Gb/s ring transport infrastructure when 10 Gb/s Ethernet data connectivity is already in the LAN backbone hardly seems worthwhile. It is simply too little too late. What is needed is a new approach that avoids the high cost of optical-electrical-optical (OEO) regeneration at every node while preserving the network architecture, topologies and operational models that have served us well, and are deeply entrenched in all our business and operating procedures.

Over the last three years, ROADM has transitioned from being too expensive for a metro environment to offering a compelling business case that is turning the heads of many metro-regional service providers. The emergence of less expensive optical components engineered specifically for metro applications along with advances in optical layer management have reduced the cost of implementing a dynamic OADM optical infrastructure by an order of magnitude. Thus, the way has been paved for a logical evolution to the ROADM optical layer.

The latest evolution of the ROADM infrastructure addresses the issues of growth, cost and service delivery requirements in the metro-regional network. The ROADM approach is designed to: 

• eliminate stranded bandwidth in metro-regional Sonet/SDH networks;

• speed service provisioning by decoupling it from network design;

• support today’s point-to-point and ring topologies

• provide a platform to evolve to more efficient mesh network topologies;

• support flexible optical layer protection on a per service basis;

• deliver lower first-in network cost and offer pay-as-you-grow scalability;

• eliminate disruption to existing services during growth; and

• support real-time optical monitoring and wavelength path management capabilities for service assurance and fault sectionalization.

Comparing Architectures

Before considering architectural tradeoffs in the design of a ROADM node, first let’s consider some common requirements. No matter what the ROADM architecture, the large number of cascaded optical amplifiers in the network must effectively deal with potentially large optical power transients to prevent surviving traffic from being affected when wavelengths are added or dropped. Sub-wavelength multiplexing must be accomplished using integrated Sonet ADM/DCS functionality or more advanced multiplexing techniques based on virtual concatenation and generic framing procedures (GFP). Finally, the dynamic behavior of ROADM nodes requires a control plane that will correctly configure the node and perform essential signaling and switching functions to implement protection and restoration mechanisms.

Where ROADM architectures begin to differ is in the cost and operational impact of the initial installation and the capacity growth scenarios thereafter; the optical penalties such as loss and dispersion that the architecture implies; the fault characteristics of the architecture, in terms of both its inherent reliability and the ability to isolate faults when they do occur.

Finally the technologies with which an architecture can be implemented, and how well the implementation can adapt to changes in technology and network architecture can dramatically affect the “cost curve” that a given architecture will ride, either closing the door to future cost reductions, or being able to ride the “optical Moore’s law” of lower costs and higher performance.

Three potential ROADM architectures can be compared in each of these dimensions:

• a basic demux/switch/mux (DSM)-based ROADM

• a banded DSM-based ROADM and

• a broadcast and select-based ROADM.

Basic DSM-based ROADM

In the basic DSM-based approach, the full DWDM spectrum is demultiplexed, switched and then re-multiplexed as a monolithic block of wavelengths. The use of electronically-variable optical amplifiers (eVOAs) on each wavelength through-path enables power adjustments to accommodate the variable per-wavelength losses that naturally occur in fiber and optical amplifiers as well as those induced by using multiple, parallel through-paths.

The multiplicity of optical taps can be used to monitor per-wavelength power levels for correctly setting the eVOAs. While this approach is very straightforward and offers excellent flexibility, the architecture suffers from several disadvantages in cost, performance and availability.

A key problem with basic DSM is that the entire ROADM fabric—including multiplexers, the multi-port optical switch, and remultiplexers—must be deployed and paid for before lighting the first service. This high start-up cost precludes the cost-effectiveness of this architecture in all but the densest and highest-growth network and market environments.

DSM architectures are typically implemented using array wave guide (AWG) filter technology along with MEMS switching to keep costs as low as possible. While AWG performance is improving, the attenuation and spectral narrowing effects of all the cascaded filters can both increase loss and create very tight constraints on wavelength and filter center frequency accuracy. The higher losses lead to mandatory use of egress amplifiers, resulting in more optical noise, more cost, and a lower overall cost/performance ratio.

The switching function of the DSM represents a single-point-of-failure in the node design, resulting in the replacement of the entire switching element should a failure occur at any one of the many fiber connections. All connections are lost while replacing the switching module. Although reliability can be improved with the use of a redundant optical switch fabric, more loss is added and start-up costs are again increased.

Banded DSM-based ROADM

As an effort to avoid the high start-up cost of the DSM architecture, the spectrum of the DSM architecture was divided into “bands” and the single large optical switch was broken down into smaller elements—one per band. Although overall DSM performance and cost effectiveness can be improved with carefully chosen filter structures, such as low-dispersion interleavers, there are many possible failure points and some key flaws to flexibility.

As with pure DSM, the complex through-path creates more potential points of failure while increasing the number of connections as the filtering and switching elements are distributed. Although the ability exists to localize a failure to just one band, without duplicating the switch it still represents a single point of failure for any traffic in the band—with both through and add/drop traffic. Fault isolation also becomes more difficult with the addition of the various filter stages.

The key flaw in the banded DSM architecture is inherent in its use of wavelength banding, which can create bandwidth stranding. To reduce first-in cost, the DSM bands are typically installed incrementally as add/drop capacity is required at a given node and the undeployed bands can be “glassed through” at lower cost. The issue is that the full spectrum of wavelengths is unavailable. For example, only eight wavelengths are available in a 32-wavelength/four-band implementation.

Architectures that keep module interconnection simple and minimal offer the greatest opportunity to follow the 'optical Moore’s law' and provide the greatest future cost reductions.

If an additional ninth wavelength is needed, the choice becomes upgrading the end points or upgrading the end points and all the nodes in between. Upgrading only the two end points—the less costly approach—guarantees service outages since the intermediate nodes will, in turn, require upgrading. This is unacceptable in core transport networks.

Demand for capacity with limited usable wavelengths make “pay-as-you-grow” deployment virtually impossible. Though initially appealing for cost and performance, the banded DSM architecture fails due to conflicting requirements for flexible wavelength coloring and zero disruptions during capacity growth.

Finally the problem that both banded and unbanded DSM suffer from is that they “split open” the optical through-path at its widest point, when fully demultiplexed. These fully demultiplexed signals must be connected between modules, and as anyone who has watched the evolution of silicon technology can attest, packaging and interconnect costs very rapidly come to dominate the total cost for modules. Architectures that keep module interconnection simple and minimal offer the greatest opportunity to follow the “optical Moore’s law” and provide the greatest future cost reductions.

Broadcast and select-based ROADM

In the broadcast and select ROADM architecture, there is just one through-path, and the entire spectrum is always available for add/drop at any node. The full wavelength spectrum is split at node ingress and sent to both add/drop filters as well as towards the node egress. A combination wavelength blocker/dynamic channel equalizer (DCE) selectively blocks wavelengths that are used for add/drop (so they can be reused at this or other nodes) and provides the per-wavelength power control function, as opposed to using individual eVOAs in the DSM architectures.

Since all capacity upgrades for a broadcast and select architecture occur off the ring and can be performed incrementally at only the source or destination necessary, incremental revenues will likely offset incremental cost. The reduction in module interconnection also dramatically decreases packaging and interconnect costs—a key cost-reducing area given today’s technology limitations.

The broadcast and select design also employs a “keep it simple” approach to ensuring the through-path is reliable. Fewer points of failure equate to better reliability.

The coup-de-grace for this architecture is in offering the best features of the two DSM architectures, but without the disadvantages. Any wavelength from the entire spectrum can be used for add/drop traffic, similar to the pure DSM, yet the add/drop filters can be installed incrementally, as in banded DSM. There are no inherent service disruptions, and technologies are available for both DCE and add/drop filtering that render this architecture more adaptable to advances in component technology.

Wavelength tracking

Although a clear winner in comparison with either DSM architectures, the potential Achilles heel of the broadcast and select ROADM is that the through-path is not “opened up” to allow per-wavelength monitoring for fault isolation and controlling DCE power levels. Another challenge is that the individual through-wavelengths are not “broken out” for fault isolation and power management. The best approach to solving these wavelength management issues is to apply a technique know as wavelength tracking technology.

Wavelength tracking applies a unique optical signature to each wavelength that enters the OADM layer and used this signature to track, monitor, and report on each wavelength. This technology even distinguishes between multiple instances of the same color traveling through the network. By tracing the end-to-end path and power level of each wavelength, management can be accomplished without the need for OEO conversions—reducing both complexity and cost.

As networks migrate from simple point-to-point DWDM to optical rings with OADMs, metro-regional topologies must also migrate onto reconfigurable, manageable, and cost-effective architectures. A broadcast and select-based ROADM architecture, coupled with wavelength tracking technology, enables service providers to design and build flexible, scalable, cost effective and manageable optical networks— enabling them to both see and control the light.

---

Ben Bacque (bbacque@tropicnetworks.com) is vice president of product line management and Dan Oprea (doprea@tropicnetworks.com) is vice president of system architecture at Tropic Networks, Ottawa, ON, Canada.

Visit Tropic Networks online.