News & Analysis
What DWDM's lambda-banded approach means for carriers
Lars Friedrich and Henning Hinderth¼r
1/29/2004 5:26 PM EST
All signs indicate that fully meshed patterns and dynamic changes are the future for traffic in metropolitan optical networks. Dense-wavelength-division multiplexing (DWDM) technologies are evolving to reflect the trend by handling single wavelengths flexibly and at a reasonable cost.
Today's DWDM filter stages and optical amplifiers are designed to cover as many channels per module as possible. Protection schemes, meanwhile, are adapted from the synchronous optical network/synchronous digital hierarchy (Sonet/SDH) world; they deal with single wavelengths, and the associated costs are tremendous.
A solution is emerging for this inherent conflict: a two-stage DWDM filter that aggregates multiple wavelengths to a l band and then allocates individual bands for individual traffic relations. For the metro carrier introducing managed offerings of enterprise-specific, high-bandwidth storage and Ethernet services, this is a promising development. The lambda-banded approach--delivering the ideal level of granularity for cost-effective routing, amplification, and protection--renders networks more flexible and slashes cost per service.
In Sonet/SDH, the entire bandwidth payload is accessible at every network node. Offering the same degree of access to bandwidth carried by wavelength channels would require deployment of DWDM filters for all channels at every node. While technically feasible, the carrier's initial investment would be prohibitive. First, the optical filter modules themselves would represent a significant amount of the total equipment cost. Second, optical amplifiers would often be required to overcome the finite insertion loss of these modules. Ultimately, the overall system budget would be significantly reduced.
As a compromise, the entire DWDM spectrum can be separated into different wavelength bands, each supporting a subset of the total wavelength count. The optical filter modules are organized into two stages. This technique, lambda banding, is illustrated in Fig. 1. A band-splitter stage separates the full spectrum into a number of bands, each carrying several channels. The second stage then separates the channels at the single-wavelength level. Isolation values are high for both filter stages. Deploying the first filter stage at each node is sufficient to enable access to the entire spectrum; the second channel-level filter stage is deployed only if and when required.
In this way, the lambda-banded approach walks the middle ground of entailing moderate initial costs, compared to a minimal solution, while offering full extensibility. The number of necessary amplifiers, optical (line) switches and other components is reduced, and traffic relations in the ring topology effectively break down to point-to-point links at the band level.
Lambda banding is increasingly prevalent in metro DWDM systems. Many vendors offer DWDM systems with 24, 32 or 64 wavelengths and typically three, four or eight wavelengths per band.
Impact on amplification
One of the key advantages of today's popular Erbium-doped fiber amplifiers (EDFAs) is their ability to amplify many channels at once, due to broad gain spectrum and an ability to achieve highly saturated output powers. This feature is useful in long-haul DWDM networks, where many channels are to be transported and little traffic is dropped at intermediate sites. In the metro space, however, with its ring topologies and meshed traffic relations, giving up part of this advantage cuts carrier costs and boosts flexibility.
Gain characteristics of conventional, broadband EDFAs saturate at high input powers; consequently, these amplifiers require all entering channels to have similar power levels. Dissimilar power levels result in reduced optical signal-to-noise ratio (OSNR) and, eventually, reduced bit-error-rate performance on the weaker channels. But in the complex, meshed traffic patterns of metro ring architectures, the power levels of individual WDM channels-or, in a banded system, of the different bands-are inevitably varied.
In a conventional system with EDFAs amplifying the entire spectrum, all channels must be leveled for the power requirement of the weakest channel. In l banding, EDFAs amplify only those bands that have dropped far enough in power level. (High isolation values are critical here again, to prevent crosstalk from adjacent channels in adjacent bands.) The lambda-banded approach-avoiding attenuation of certain channels-is more effective. While more amplifiers might have to be deployed (per band), leveling components such as variable optical attenuators are not required (per channel). Capital expenditures are, on balance, reduced.
The lambda-banded approach also delivers long-term flexibility in this area. Adding wavelengths or traffic relations among existing nodes requires only additional transponders and the second filter stage in cases where the band is completely filled or at relevant nodes. No existing traffic must be taken down, and the optical layer's physical characteristics are largely unaffected.
Sufficient protection
Today's predominant transmission technology remains Sonet/SDH. Most of the service-level agreements (SLAs) that carriers have struck with their metro customers are based on Sonet/SDH protection characteristics. Therefore, there has been significant effort to transfer the principles of Sonet/SDH protection and self-healing rings (i.e., 50 milliseconds switchover and permanent monitoring of both paths) to DWDM.
DWDM schemes such as optical unidirectional path switched ring (O-UPSR) and optical bidirectional-wavelength path switched ring (O-BLSR) protect the entire path, end-to-end per individual channel. End-to-end protection in a DWDM network means full redundancy in all network sections--redundant fiber traces, redundant filter stages, redundant active WDM channel cards. Because channel and band filter stages are separated, both filter stages and channel cards are redundant. If the working path fails, the protection channel card tributary interface activates; protection is achieved at the single-wavelength level.
In lambda banding, protection is delivered at the band level. An optical band switch between band and channel filter stages is going to do the rerouting between working and protection traces for an entire wavelength band. In the band protection scheme, one switching mechanism addresses all services of a traffic relation. Corresponding switches at both ends of a traffic relation always switch in parallel. Each corresponding pair of switches has its individual switching criteria.
The logical consequence when reducing redundancy in a network is an increase in the possibility of network downtime. In a path-protected DWDM terminal, channel cards, channel filters and band filter stages are redundant. The band-protection scheme cuts channel card and channel filter redundancy and introduces a new component-the band switch-as an additional possible single point of failure. So, the probability for service interruption in a band-protected network must increase in some way.
But metro carriers do not leave their network downtime to chance. They plan and design networks according to well-defined network availability parameters such as mean time between failure (MTBF) value of transmission equipment, mean time to repair (MTTR) value of transmission equipment and downtime of fiber per year and per kilometer.
Typical SLAs among numerous metro carriers guarantee availabilities ranging from 98.000 percent to 99.990 percent. For the four-node ring scenario, network availability is calculated at 99.997 percent for a path-protected ring with a 100-km circumference (based on standard, DWDM metro equipment values of four hours for MTTR and fiber downtime of 0.563 hours per year and kilometer). An equivalent calculation for a DWDM ring network featuring the lambda-banded approach leads to an overall availability of 99.993 percent. This means the concept of band protection meets the network availability requirements that metro carriers must achieve when offering protected premium services to end customers.
On the other hand, protection in the new lambda-banded approach offers the opportunity to considerably reduce capital expenditures. Because configuration of channel cards and channel filters is not redundant, capital expenditures per service can be reduced by between 25 percent and 45 percent. This cost savings depends on the number of dropped wavelength channels per optical add/drop multiplexer (OADM) node.
Lars Friedrich and Henning Hinderth¼r are product-line managers for the FSP 2000/3000 WDM platforms at ADVA Optical Networking Inc. (Mahwah, N.J.).


See related chart
See related chart
