A fresh look at delivering dynamic optical services

Delivering dynamic services is not quite as simple as it sounds. A unified optical control plane can help fulfill the promises made during telecom's boom years

Quite a few promises were made during the telecom boom of the late 1990s—some of which were kept, others which weren't. There were promises of stock options riches, of revolutionary technology, of new service opportunities. One of these promises was the emergence of dynamic optical services—like bandwidth on demand. Enabled by new optical technologies such as dense wave division multiplexing (DWDM) and optical switching, the technical and financial benefits of these services were touted as clear, compelling and revolutionary.

Looking at this promise from the perspective of incumbent service provider networks, for example, it becomes clear that the delivery of dynamic services is not quite as simple as the marketing hype would make you believe. Service provider networks consist of many different layers, sub-networks, support systems and equipment from many different vendors. Even considering the incredible technical advances achieved in individual network elements—such as intelligent optical switches—service provider networks still require a common control plane across network elements (edge, core; optical, electrical) to enable the service side devices such as IP routers to tap into the increased capacity and flexibility of the optical layer.

To address this challenge, a number of organizations have been developing signaling and routing protocols to create a unified optical control plane. In reality, dynamic optical equipment has been around for years, and several control plane initiatives have come and gone, and moved in and out of favor. From the perspective of a service provider, the number of organizations, protocols, and architecture choices only adds to the complexity and the inherent challenges of introducing next generation technology into existing network architectures.

The desire to be able to connect client layer devices such as IP routers originally stemmed from the promise of delivering new dynamic services from the optical layer. As new optical switches and DWDM technologies brought increasing and previously unheard-of capacity and dynamic capabilities to the optical layer, service providers and equipment vendors both recognized the value of extending the capabilities of this new dynamic optical layer to accommodate the data-centric traffic of devices such as IP routers.

More recently, market conditions have driven service providers to take a fresh look at optical control plane initiatives based on a more pragmatic, rational benefit: reducing operational expense. In contrast to the traditional, manual provisioning processes and time-intensive manual operations, a unified optical control plane simplifies many aspects of OAM&P (operations, administration, maintenance and provisioning), enabling service providers to greatly reduce operational expenses. Furthermore, the adoption of a distributed, IP-based optical control plane increases network scalability and the ability to extend it across multiple layers of the network.

The combination of these immediate, practical benefits and the still viable long-term promise of dynamic optical services has been driving development of optical control plane technologies, despite current economic and industry conditions. In fact, progress has been considerable, as evidenced by the number of public demonstrations, extensive multivendor interoperability tests and the number of leading-edge equipment providers that are currently delivering support for some of these protocols on their generally available products.

While an overlay approach presents a simpler solution, the peer model offers potentially greater cost savings, service opportunities and efficiency gains in the long-term.

Too Many Cooks

Since work began on optical control plane standards, many groups have advanced their own interpretations and approaches, all with the goal of developing a unified optical control plane. In fact, the number of disparate organizations has likely had a “too many cooks in the kitchen” effect, contributing to the confusion surrounding control plane standards and delaying market acceptance.

Some of the organizations that have worked on developing standards include the (now defunct) ODSI, the ITU-T, the IETF and the OIF. While the IETF’s GMPLS initiatives and the OIF’s UNI/NNI approach have been attracting the most attention, it’s important to note that organizations like ODSI were not without their benefits. They ultimately provided a solid technical foundation and were a market catalyst for the standard initiatives making progress today.

Generalized Multi-Protocol Label Switching (GMPLS), the IETF’s approach to optical control plane, extends proven MPLS signaling and routing protocols to the optical domain. GMPLS permits the dynamic inter-working of SONET/SDH, wavelength, fiber and port switched networks. GMPLS creates a unified optical control plane between these elements, within and beyond the boundaries of a single network, layer, or vendor, enabling service providers to create, manage and troubleshoot services end-to-end over a variety of network elements.

The OIF optical control plane approach is to work in a participative process with its members, to define implementation agreements that foster the deployment of interoperable optical networking products for data switching and routing. The OIF user-to-network interface (UNI) and network-to-network interface (NNI) provide the service control interface between the transport network and client equipment, and between transport networks, respectively.

UNI signaling invokes services that the transport network offers to clients, enabling dynamic interconnection of client layers like IP, ATM and SONET. The NNI standard supports a service control interface between multi-vendor elements within a single transport network, between multiple transport networks within a single service provider, and between service provider networks. Together, UNI and NNI create an optical control plane between these elements which service providers can use over a variety of network elements.

The ITU-T is developing two protocol-independent control plane framework models: the general automatic switched transport network (G.ASTN) and the more specific automatic switched optical network (G.ASON) recommendations. G.ASTN describes client and technology independent network level control plane requirements for automatically switched transport networks, and G.ASON describes architecture for automatically switched optical networks. Since the ASTN/ASON model is protocol independent, GMPLS, UNI, and other protocols can be mapped to fit the framework.

Sorting Out the Choices

In addition to the organizations described above, service providers also have varying implementation models from which to choose (Figure 1). In the overlay model, the control planes of optical layer devices and the client devices are kept separate. Client devices have no knowledge of the optical topology or resource information. In the peer model implementation, all of the devices in the network—IP routers and optical and everything else—function as “peers.” All devices share topology information, and the devices function as a cohesive network whole.

Figure 1 Architectural models of control plane implementation

Clearly both architectural approaches each have their respective merits, or there wouldn’t be two approaches in the first place. The overlay model fits most closely with service providers’ present mode of operations, because it keeps the IP and optical domains functionally separate, which is currently the way most service provider organizations are arranged. For this reason, many speculate that overlay will be the first implementation seen in production networks.

While an overlay approach presents a simpler solution, the peer model offers potentially greater cost savings, service opportunities and efficiency gains in the long-term. By virtually collapsing the layers that exist within a service provider network, management software and provisioning process, a peer implementation greatly simplifies the network from a management perspective. This enables service providers to recognize cost savings and more service opportunity than even in the overlay model.

Considering the merits of each, the architectural model adopted should depend on the specific needs of the individual service provider. In support of these requirements, many equipment vendors have taken an architecturally neutral approach, enabling service providers to deploy the architectural model of their choice. By taking a neutral stance, these equipment vendors also enable hybrid models, wherein part of the network could use a peer implementation, and other parts overlay.

Which Protocol?

Betting on which organization’s approach—GMPLS, UNI/NNI, ASTN/ASON—will be adopted first and most prolifically is considerably more difficult. Previously, most argued that since OIF UNI/NNI was focused on the overlay model, these protocols would be the first to find their way into service provider networks. Now however, GMPLS also incorporates an overlay approach, and in fact uses many of the same protocols (for example, RSVP-TE) as OIF UNI and is more closely aligned with IETF standards that IP routers follow.

One thing that is a little clearer is the type of protocols that will eventually find their way into service provider networks. Given the fact that the vast majority of traffic-driving network evolution is IP-based, IP-based signaling—such as RSVP-TE—and routing protocols—such as OSPF-TE—are the likely frontrunners.

Protocols that extend ATM-based practices are discordant with the overwhelming trend toward IP-based, data-centric communications. These protocols also run into problems with scaling in mesh environments, requiring the introduction of sub-networks and various hierarchies—which is contradictory with the goal of network simplification and management layer consolidation.

Practical Considerations

Despite the expense and complexity of maintaining proprietary provisioning systems, service providers are not going to make an overnight transition to a unified optical control plane. The implementation of a unified optical control plane will be a practical, gradual transition driven by specific business needs and end-customer requests, and any optical control plane introduction must integrate with, or, at a minimum, coexist with, existing operational procedures and management systems. And regardless of how popular any particular standard becomes, no new technology ever completely replaces the older technology. (For example, the current Boston-area Yellow Pages lists no fewer than 30 entries under “Typewriters” and “Typewriter services.”)

Taking all of these factors into consideration, it becomes clear that for control plane standards to ever reach the point of widespread adoption, intelligent optical switches must be able to support a variety of emerging control plane standards, as well as existing SONET/SDH provisioning processes and protection mechanisms. By enabling the flexible deployment of control plane standards in conjunction with traditional processes, intelligent optical switches can underpin the transition to an optical control plane.

Intelligent optical switches can be introduced transparently into existing networks, to consolidate the functions of traditional equipment such as broadband digital cross-connect systems (BB-DCS) and add/drop multiplexers (ADM). In this function, the switch functions transparently as both a BB-DCS and an ADM, and must support the features and functionality of both. The focus is on operational transparency, and the switch needs to support traditional protection mechanisms such as BLSR/MS-SPRing and UPSR/SNCP.

In this first phase, the intelligent capability (routing and signaling capability) of these new elements will likely not be used. From an operational perspective, these elements will be integrated into the present mode of operation (such as static, manual provisioning), and integrated into the appropriate elements of the OSS (inventory, management and provisioning).

With the intelligent optical switches in place, service providers can begin to evaluate the introduction of the optical control plane of their choice. The intelligent optical switches will selectively implement the optical control plane, via software, on a per-port basis.

During this transition, the optical switches will support both static and dynamic provisioning, as well as traditional protection mechanisms such as BLSR/MS-SPRing and advanced mesh protection schemes. This approach enables service providers to implement GMPLS and/or UNI/NNI selectively, without disruption to existing services or revenue steams. As mentioned earlier, in the near to mid-term, it is likely that an overlay model will be adopted for isolation and security reasons.

Over time, as service providers gain experience and confidence with their optical control plane implementation they may begin to implement an optical control plane throughout their network to capitalize on market opportunities and operational benefits.

It should be noted that there will always be a place for static, manually provisioned services and most networks will have to support many different co-existing services. With a flexible infrastructure composed of intelligent optical switches, there is no need to uproot existing services, and customers who prefer or require traditional services and protection mechanisms can be accommodated. Figure 2 below illustrates a possible end-goal control plane implementation strategy in which optical switches enable a combination of dynamic and static service offerings.

Figure 2 Possible control plane implementation: A mix of static and dynamic

While the dominant control plane standard of the future is far from decided, the current variations and development approaches should not be seen as a negative. In fact, the variety is likely to drive development and encourage cooperation between equipment vendors. In the meantime, service providers can begin to evaluate intelligent optical switches that can support multiple control plane standards simultaneously, while maintaining support for traditional provisioning processes and protection schemes. With intelligent optical switches, service providers can evaluate these protocols today, and can even selectively deploy control plane standards like GMPLS and UNI/NNI to support specific services.

Ultimately adoption of a unified optical control plane will be driven by a combination of the immediate operational benefits and customers’ need for flexible, dynamic services. With intelligent optical switches and the advances leading to the adoption of a unified optical control plane, dynamic services like bandwidth on demand are a promise of the telecom boom that has not yet been broken—only delayed.

---

Tom DiMicelli is Senior Product Manager for Sycamore Networks, Inc., Chelmsford, MA. He can be reached at thomas.dimicelli@sycamorenet.com.

Visit Sycamore Networks online.