Photonics a gamble that will pay

The introduction of transparent photonic switching into communications networks is sure to provide savings in several areas. Photonic switching is likely to supplement current deployment initiatives for STS-1 grooming resources. Using photonic switching, designers will be able to combine transparent trunking, transponder and optical-electrical-optical (OEO) conversion optimization, banded wavelength routing and trunk grooming to provide the critical bottom-line business impact.

Our baseline model assumes ongoing buildout of an STS-1-based grooming switch infrastructure. With this initiative in process, why would any provider consider installation of a photonic switch? The bad economy and drop in demand for high-speed optical connections means service providers now consider grooming switches a necessity. Since switching requirements at the optical level (2.5 Gbits/second and 10 Gbits/s) are small, service providers can offset the growth/scalability argument of a photonic switch with the capital-expenditure penalty of a few 2.5- and/or 10-Gbit/s ports. In other words, they trade functionality for price. They are willing to pay more by running 2.5- and 10-Gbit/s transit traffic less efficiently through the grooming switch. However, a slight glitch has emerged: Developers are introducing grooming switches as de facto mesh core solution switches, even though these devices may have a price penalty if the network starts growing again. And networks will grow again at some point; they'll have to, to absorb the traffic when the economy picks up again. Conversely, if one switch or technology supports multiple functions in the network, the integration with the BSS/OSS systems becomes a lot easier.

So, when does photonic switching prove itself? Almost immediately. Inserting a photonic switch in a DWDM hub results in immediate capital-expenditure benefits, even if the amount of drop traffic is low. Mature microelectromechanical systems design and high-yield manufacturing, combined with new optical packaging, have created reliable, low-cost photonic switches that are one-quarter of the OC-48 per-port price of electronic fabric systems now being deployed; for OC-192, that factor becomes one-sixteenth. This demonstrates the economic advantage of switching at the optical layer.

Distinct advantage

While there are distinct advantages to and a definite need for STS-1 grooming switches, deploying them as core network trunking switches carries a cost inherent to the switch architecture in general, and the lack of "transparent" ports in particular. Consequently, advances in line systems, such as the migration to 10 Gbits/s, or more-efficient protocol architectures, require wholesale upgrades of the I/O ports, if not the switching systems themselves. On a more microscopic level, we observe that there are two aspects to the problem of providing efficient and cost-effective switching solutions. The first is the need to maximize the number of revenue-generating service ports. The second is to make efficient use of the new higher-speed transport systems.

It is difficult to arrive at an optimal solution, in view of the constantly evolving traffic dynamics at any one switching hub. OEO switches have fixed and rigid switching capacities, and the total capacity must be divided between the needs of the line systems and the revenue ports with the above objectives in mind. For example, to interface to a 10-Gbit/s line system, line-associated switch capacity has to be allocated in discrete 10-Gbit/s blocks. This potentially prevents maximization of the number of revenue-generating service ports that the switch can provision. Consider the case of pass-through, or nongroomed, and other marginally groomed service traffic. Pass-through traffic is typically aggregated high-speed traffic, so the transit of such connections through an OEO optical switch will consume capacity that potentially could be used to provision services. The same applies to marginally groomed-essentially transit-traffic that requires minimal grooming between transit ports.

Data interfaces are migrating to ever-higher speeds, approaching those of the line systems themselves. Past solutions meant provisioning these circuits directly on the line systems via patch panels. This is acceptable when the number of circuits is low, but becomes unmanageable when the number of circuits increases. Even with low circuit counts, performance guarantees and liabilities imposed by service-level agreements demand that circuits be provisioned on manageable network elements, the most appropriate being optical switches. However, provisioning of such circuits on OEO switches further burdens the already constrained capacity. The problems are not insurmountable, however. Developers have produced new network elements to handle these trouble spots. The availability of reliable and low-cost photonic switches, for example, offers practical solutions that address technical, operational and business aspects of the problems at hand.

It is important to note that the photonic switch is not an alternative to the OEO grooming switch, but rather is a coexisting platform that provides relief for particular problems. This coexistence works from a business standpoint as well as an operational one. As mentioned earlier, inserting a photonic switch in a DWDM hub can generate immediate capital-expenditure benefits, even if the amount of drop traffic is very low. Operations could be disrupted, though. Beyond having to manage two network elements, both switches have to support coordinated provisioning, which requires an intimate integration through a common control plane like generalized multiprotocol label switching(G-MPLS), which is still in its infancy.

In its most recent report, "G-MPLS and the Optical Control Plane: An Analysis of Profitability and Performance in Optical Networks," Pioneer Consulting developed a financial model that quantified the financial benefits of G-MPLS for network operators. Pioneer said G-MPLS can provide sizeable decreases in operational and capital expense accompanied by material increases in revenue. The report noted, however, that deployment of G-MPLS won't occur until 2004 or 2005. Until then, service providers may deploy a proprietary intelligent control-plane solution and then migrate to G-MPLS several years later.

Hybrid switches

There are two flavors of hybrid OEO/OOO switching: a single-vendor solution that integrates a photonic-switching module with an OEO grooming switch and a multivendor solution in which a photonic switch from vendor A supplements a grooming switch from vendor B. The latter requires a standardized control plane like G-MPLS, while the former can start with a proprietary control plane and then migrate. The integration exercise doesn't end there, however. DWDM systems are a key network element in optical network transport hubs, and they can play a significant role in reducing capital and operating expenses.

Greater gains

For example, once a service provider decides to deploy photonic switches, the provider will begin to explore how it can obtain even greater gains by removing expensive components like transponders from the optical paths. Additional functionality can be gained if the provider bundles wavelength traffic at the edges of the optical cloud, to provision transit routes that reduce the interface count at hub points by a factor of 5 to 10 times. Capital and operational savings ensue because the complexity of the handoff has been reduced. Finally, in the specific application of acting as a high-bandwidth trunk manager at the edge of the all-optical cloud, the photonic switch can extend its value by providing subwavelength grooming at the OC-48 level.

Depending on the network's complexity and load, deploying a photonic switch for subwavelength grooming at the OC-48 level could save between 10 percent and 40 percent of the ports used for trunking on the STS-1 grooming switch. Providers could, in turn, redeploy the ports so that they generate revenue for customers, a more cost-effective use of capital. In many cases, this trunking application becomes crucial as networks expand. Often STS-1 grooming switches are port- and switch-core-limited. One popular product in this class can terminate no more than 32 bidirectional OC-192 connections. To extend life and usability, a photonic switch could be used to take over the trunking function from the grooming switch. This would result in the best combination of high-bandwidth, high port-count photonic switching with "pay as you grow" electrical grooming resources.

It's possible to perform hybrid switching by combining equipment from three vendors, each one supplying a single piece of the solution. Choosing one vendor to supply a combined OEO/OOO switch and another for the DWDM system reduces the number of vendors. Alternatively, one vendor could offer a combined transport/switch solution (an optical add/drop switch). Finally, a single-vendor solution requires integration of all three network elements in the hub. These options require a standard control plane to support end-to-end efficiency and fast service activation.

While it seems obvious that the most pragmatic solution would be to deal with one vendor where OSS/BSS integration is limited to a single network, there is a glitch: This solution requires a significant network overbuild. The integration hurdles are bigger than with alternative approaches, since the new network element needs to interact with two existing elements that may not support the latest control plane technology. The good news in this context is that several software vendors have developed middleware that bridges the gap between the new, G-MPLS-based control plane and the legacy Sonet and Osmine provisioning solutions.

Optical advantage

Clearly, there are economic advantages to switching at the optical layer. Reliable and low-cost photonic switches are available today at one-fourth of the OC-48 per-port price of current electronic fabric systems. For OC-192, that factor becomes one-sixteenth of the price of current fabric systems.

The immediate benefit of a photonic-switch deployment is the dramatic reduction in OC-192 trunking ports on the OEO switch. This decrease stems from the transparency of the photonic switch, allowing all pass-through traffic to be offloaded from the OEO switch.

It is relatively easy to demonstrate the financial benefits of hybrid switching. However, introducing this architecture could disrupt operations, obviously a big drawback. In a multivendor environment, multiple network elements need to support coordinated provisioning; that requires intimate integration through a common control plane like G-MPLS, which is still relatively undeveloped.

Conversely, this approach allows an easier migration strategy by building on an existing network. Highly integrated single-vendor solutions do not require a standard control plane but will require a significant network overbuild.

A complete vesion of this article was presented at the 2002 Communications Design Conference.

See related chart

See related chart