Maximum reliability and performance for data-optimized backbone networks

John Nikolopoulos

This year and next will see the deployment of several large optical dense wavelength-division multiplexed (dwdm) rings with an OC-192 (10-Gbit/ sec) 4-fiber Synchronous Optical Network (sonet) bidirectional line-switched ring (blsr) configuration mapped onto a 2-fiber plant using bidirectional line-amplified technology. The blsr architecture--a proven and highly cost-effective technology in large metropolitan networks--is being scaled up to support the first national backbones of the new millenium. Qwest Communications International Inc. led the way this past June with the launch of the world`s first dwdm OC-192 4-fiber ring system in its high-capacity fiber network.

The deployment of such rings is one of the effects of the deregulation of interexchange traffic. National network operators are looking for a robust technology that can cost-effectively handle the demanding backbone networking requirements for a wide variety of services (see Fig. 1). And the best way to obtain simple, flexible, and self-healing long-haul architectures is to leverage a very versatile and resilient platform--the sonet standard.

Leveraging sonet

The dwdm OC-192 4-fiber blsr is indeed the leading edge in deployed transport technology (see Fig. 2). The new data-optimized national backbone networks that will use this topology will require equipment that supports the scaling of sonet technology in six significant ways:

ring circumferences greater than the 1200-km reference model

bit-error-rate (ber) performance better than traditional 10-12 levels

ring add/drop multiplexing (adm) support greater than the traditional 16-node limit

mapping of a logical 4-fiber blsr onto a 2-fiber plant, thus saving fibers

protection channel access for greater traffic-carrying capacity

global service mix support via integration of sonet and Synchronas Digital Hierarchy (sdh).

Compared to its 2-fiber counterparts, OC-192 4-fiber blsr enhances the ring architecture in two important ways. It doubles the traffic-handling capacity and offers both ring- and span-protection switching modes; the latter offers self-healing during multiple fault conditions and, hence, a high degree of survivability. The OC-192 4-fiber blsr capability is critical for long-haul networks that involve the traversal of multiple maintenance boundaries. With the span switching inherent in a 4-fiber blsr, a network operator can segregate the maintenance activities of the whole network for better coordination and to eliminate the chance of a costly redundant network-maintenance action.

Traditional sonet/sdh network elements have evolved considerably, integrating conventional crossconnect and adm capabilities into a sophisticated bandwidth-management vehicle that offers operational cost savings and transmission reliability. Integration of some layer 2 and 3 functionality is on its way as well.

Given all the technologies we have today, the dwdm OC-192 4-fiber blsr architecture delivers the lowest cost per bit-mile. More importantly, it shows that sonet technology can scale up to deliver simple robustness at a lower cost, and also interface with optical layer technology in an integrated and efficient manner. Some say that sonet is simply a legacy technology pushing more capacity onto the network. In fact, sonet is showing that it has the flexibility to be scaled up to deliver powerful efficiencies and economies of scale in long-haul networks--without sacrificing reliability and performance.

Moreover, sonet does not inhibit the deployment of open optical interfaces. On the contrary, the rich sonet operations, administration, maintenance, and provisioning (oam&p) capability has become an integral part of such interfaces. Finally, advanced sonet survivability provides a solid platform to support Internet protocol/Asynchronous Transfer Mode- (IP/atm-) based virtual ring long-haul applications.

Scaling the standards to match emerging long-haul requirements, including support for larger sonet ring circumferences and rings capable of supporting more than 16 nodes, is merely a new application for a mature technology. Recently, it has been shown that such rings can create extremely cost-competitive and operationally simple networks without compromising such objectives as high service-carriage quality, system robustness, and reliability.1

Another example of scaling is the ber. With the forward error correction (fec) capability embedded within the OC-192 sonet overhead, the ber can be improved from 10-12 to typically 10-20 for a given optical reach. Alternatively, fec can be used to improve optical reach by at least several decibels for a given ber level. Perhaps most importantly, fec enables proactive maintenance via access to both the raw (before fec) and corrected (after fec) ber statistics. This access allows operators to detect marginal fiber plant/equipment degradations and effect proactive maintenance if necessary.

Scaling flexibility is also observed in the 2-fiber bidirectional line amplifier implementation associated with the dwdm OC-192 4-fiber blsr architecture. Some of the key advantages of this implementation in long-haul networks include fiber and equipment savings on initial deployment, improved reliability because of fewer active elements at a lower cost (fewer spares), and enhanced network flexibility and simplified stacked ring networks.

The dwdm OC-192 4-fiber blsr nodes being deployed today feature full nonblocking sts-1 (52-Mbit/sec) bandwidth management granularity that allows any eastbound or westbound sts-1 time slot to be assigned to any tributary port exactly as desired. This drastically reduces, if not eliminates, the need for cumbersome back-hauling of traffic to a hub site for external grooming and crossconnection, since the node has embedded crossconnect/adm functionality. Moreover, with full 10-Gbit/sec capacity from node to node, fewer overlaid rings--and thus fewer hard-to-manage inter-ring hand-offs--are required when compared to overlaid OC-48 (2.5-Gbit/sec) ring architectures (see Fig. 3).

Operators can increase their flexibility and efficiency through the deployment of advanced bandwidth management capabilities. In the long-haul network, highly survivable ring interconnections may be required as traffic passes from one ring to another. A dwdm OC-192 platform with a matched node inter-ring gateway configuration can provide this end-to-end reliability. Duplicate, geographically diverse primary/secondary gateways pass services from an OC-192 4-fiber blsr to a neighboring ring. Traffic from the secondary gateway is automatically selected in the event of a problem with the primary gateway.

The support of subtending rings off of a dwdm OC-192 4-fiber blsr represents another example of how flexible bandwidth management can streamline operations and reduce external crossconnect ports and the equipment footprint. In this application, a service passes in from one tributary and passes out through another without grooming from an external crossconnect.

Returning to the notion of exploiting sonet topology, an OC-192 platform can be configured to exploit the normally idle protection bandwidth incorporated into most sonet configurations. Although this bandwidth must be reserved for survivability, that doesn`t mean it can`t be used for unprotected or extra traffic. Both the protection channel access (pca) and matched nodes discussed above provide flexibility in tariffing and grade-of-service offerings. For example, for some non-mission-critical data applications, pca may provide the transport pipe with restoration (with its less stringent requirements in this case) pushed to the edge of the network (i.gif., layer 2 or 3).

Another key requirement of today`s long-haul networks is the ability of network elements to function efficiently as global platforms for different service types. Given all of the above benefits, it is clear that the dwdm OC-192 platform must perform as a multifaceted global platform, not just as a "big bandwidth box." It must support a wide service mix at the tributary level, including OC-12/12c, OC-48/48c, and their sdh equivalents. The dwdm OC-192 equipment being deployed in the new national networks supports such advanced capabilities. dwdm OC-192 4-fiber blsr takes sonet to a new level of service flexibility and operational efficiencies in long-haul networks, but always within the proven sonet/sdh framework.

Integrating the optical layer

If there is one thing to be learned from all the talk about the growing range of optical network elements with embedded sonet interfaces, it is that sonet is here to stay for some years to come. Integrating the sonet OC-192 electrical layer with the associated optical layer so that they are deployed in a complementary and efficient manner is the present-day and near-term objective of network planners.

Multiplexing can be performed in the electrical layer using a time-division multiplexing (tdm) sonet system, or in the optical layer using dwdm. On a dwdm OC-192 4-fiber blsr platform, capacity is driven to the limits of current tdm offerings (i.gif., 10 Gbits/sec) with a high degree of mixed-service aggregation (OC-12/12c, OC-48/48c, etc.). This 10-Gbit/sec signal is then scaled up using dwdm. Basically, the underlying optical layer supporting a 10-Gbit/sec platform serves to deliver capacity relief by multiplexing multiple sonet systems over a single fiber. Systems capable of multiplexing 16 wavelengths at OC-192 (for a 160-Gbit/sec capacity) are available today and are readily scaleable to 32-wavelength systems (a 320-Gbit/sec capacity).

Both layers can be used in the same network to handle transport functions for optimum network bandwidth efficiencies. Although the electrical layer is well managed via sonet oam&p, achieving comparable levels of control at the optical layer requires a bidirectional optical service channel (osc) integrated into the optical amplifiers to provide "regenerator-like" operations access at remote optical-line amplifier sites. Like a sonet data communications overhead channel, the osc supports all network-management functions, including alarm reporting, provisioning, fault diagnostics, and remote software downloads, avoiding the expense of a costly external data communications infrastructure. Also, the integration of the network-management function across the sonet and optical layers into a single interface reduce operator training and would streamline trouble-shooting (see Fig. 4).

oam&p in the optical layer necessitates the integration of a comprehensive toolkit to manage the inherently analog optical-line parameters and in the process minimize the expense of external testing equipment. Some key features of this toolkit include power measurements for individual wavelengths and an integrated optical reflectometer for trouble-shooting faulty splices, optical connectors, and fiber-optic cable. Finally, an integrated intelligent system line-up and test system would significantly reduce the time and effort required to properly adjust the optical power levels when turning up or upgrading a dwdm/OC-192 system. This type of advanced toolkit effectively addresses the emerging needs of dwdm management in the network. The sophisticated dwdm/OC-192 equipment being deployed today already provides such an advanced toolkit.

Long-haul optical networking

Network planners are wrestling with the question of just how far we can go with optical networking to support data traffic in current build-outs. Soon IP and atm platforms will have OC-48c and OC-192c (9.6-Gbit/sec) service interfaces, which do not require multiplexing to higher bit rates. It has been suggested that the sonet layer can be eliminated. But at what cost?

S. Melle, et al., examined that question in a paper delivered at the 1998 National Fiber Optic Engineers Conference.2 The authors considered a hypothetical nationwide optical long-haul network designed around data-centric services originating from broadband IP routers and atm switches. This represents the ideal case for optical networking. They compared two state-of-the-art architectures (see Fig. 5). The first uses direct IP/atm connectivity (OC-48c) at the 40-wavelength dwdm optical layer with dedicated optical rings for protection switching; the second, IP/atm multiplexing to the sonet layer using OC-192 4-fiber blsr and connected to a point-to-point 16-wavelength dwdm system. Performance assumptions were based on results confirmed in a large sample of detailed network studies.

The study showed that using currently available technology, the optical layer doesn`t cost-effectively scale very well. As capacity increases, IP/atm over dwdm/sonet OC-192 has a 26% lower total cost versus the IP/atm-over-optical network architecture. The first reason for this is the shared protection in the sonet layer versus the dedicated protection in the dwdm/OC-48c optical architecture. By dedicating a wavelength channel around the entire ring, the dedicated optical ring architecture consumes more wavelength route-miles and incurs additional costs for optical regeneration for each optical channel around the entire ring. Conversely, shared protection via sonet is obtained at nearly no incremental cost and results in lower total network costs--lower than IP/atm over dwdm even with the elimination of the sonet layer.

And as bandwidth demand increases, 40-wavelength dwdm/OC-48c capacity exhausts faster than 16-wavelength dwdm/OC-192 system capacity, reinforcing the notion that maximizing bit rate per wavelength via tdm permits maximum use of the dwdm infrastructure and minimizes costs.

In the IP/atm-over-dwdm/OC-48c solution, OC-48c pass-throughs at intermediate nodes must be provisioned manually today. This contrasts significantly with the automatic software provisioning that accommodates such pass-throughs in dwdm/OC-192 4-fiber blsr systems. Also, the dwdm/OC-48c solution consumes a considerably greater number of wavelengths than the dwdm/OC-192 4-fiber blsr solution. Minimization of wavelength use bodes well for operational cost reductions with regard to dwdm optics sparing and general daily management of individual wavelengths. And the oam&p capabilities are not yet as robust in the optical layer as they are in the sonet layer. Lastly, the dedicated optical ring-protection architecture operates over two fibers, while the dwdm/OC-192 architecture is implemented using 4-fiber rings (which can be implemented over only 2 physical fibers using bidirectional dwdm). With 2-fiber dedicated protection optical rings, you lose span-switching capabilities, and that means no self-healing in the case of multiple faults and more complex maintenance in large-ring applications.

Optical network futures

To effectively address future data-service mixes, the evolution of technologies in both the optical and sonet/sdh layers must be given their due considerations. With atm/IP-based networking equipment offering OC-192c service interfaces in volume within two years, it is essential that the optical layers deployed today support OC-192/OC-192c transport. This will eliminate the significant costs and operational issues associated with network upgrades of dwdm systems optimized for OC-48/OC-48c transport. The introduction of OC-192c interfaces will also represent a key milestone in that, for the first time, the service interface rate will match the highest available line rate. As a result, the current OC-192/dwdm backbone target architecture optimized for OC-48/OC-48c service transport will enable the eventual smooth migration to an open interface OC-192c/dwdm-based optical network architecture. Admittedly, the successful penetration of such new interfaces will depend on several factors, including competitive pricing as well as the availability of reliable and flexible optical network technologies for appropriate bandwidth management.

Some of the exciting optical technologies currently under development and expected to become available in the next few years include optical-dedicated and shared-protection rings, optical crossconnects, and software-configurable optical add/drop multiplexers. These technologies will considerably increase the range of architectural alternatives available to network planners and will accelerate efforts towards optical-layer standardization, particularly with regard to protection-switching implementation.

sonet/sdh systems, with their well-defined and robust survivability mechanisms, contribute to a solid and reliable transport infrastructure that underpins most long-haul, high-capacity networks today. This reality must be properly considered in the development of effective strategies to address the efficient interworking between the potentially competing protection layers.

Traditional sonet/sdh network elements have undergone considerable evolution themselves, as mentioned previously, in the integration of conventional crossconnect and adm capability. The development of comparable capability within the optical layer will undergo a similar staged evolution over the next few years. Meanwhile, expanding on their current support for robust transmission of data traffic, these versatile sonet/sdh boxes are even expected to integrate some layer 2 and 3 functionality in the near-term, enabling more optimal bandwidth management of various broadband data-service mixes.

With the ever-expanding technology choices available in both the sonet/sdh and optical layers, it has become increasingly critical for national network builders to remain focused on meeting the challenges associated with their companies` overriding business objectives. These include the maximization of revenue growth, which can be facilitated through capacity scaleability and service-velocity capabilities; achieving cost containment via streamlined and simple operations (e.g., advanced bandwidth management and integrated network management); and finally, establishing effective service-differentiation strategies like ring survivability, enhanced grade-of-service offerings, and sophisticated oam&p features.

The deployment of large-ring long-haul networks increasingly based on dwdm/OC-192 4-fiber blsr technology indicates an industry consensus that this is the near-term architecture to successfully address these business concerns and simultaneously provide a platform from which to evolve to more advanced optical networking architectures. u

References

1.John Nikolopoulos, "Network Planning Considerations Associated with Large sonet Ring Deployment," paper presented at nfoec `98, Orlando, FL, September 1998.

2.Serge Melle et al., "Optical Networking Deployment Considerations in Long-Haul Networks," paper presented at nfoec `98, Orlando, FL, September 1998.

John Nikolopoulos is an optical network product-line manager and Louis-René Paré is the director of optical network product-line management at Nortel (St-Laurent, QC, Canada).