The optical networking market is staging a rebound. A combination of market forces is setting the stage for carrier and enterprise investments in optical networks to shoulder the burgeoning demand for bandwidth. That demand comes from the convergence of traffic carrying consumer and enterprise applications onto packet networks.

New consumer electronic devices--like the iPhone, Treo, Xbox 360 and Slingbox--require Internet connectivity, as do new consumer applications, including voice over IP (VoIP) from Skype, videos from YouTube, shopping on Amazon.com and eBay, as well as online financial management.

To support this consumer demand for bandwidth, carriers are expanding their residential access networks. In the US, the ILECs are encouraged in their massive, fiber-based, broadband access buildouts by FCC decisions, which protect them from having to offer wholesale access to this new infrastructure to their competitors. To support consumer demand for mobility, both carriers and municipalities are deploying 3G wireless, mesh Wi-Fi and/or WiMAX networks.

Meanwhile, voice and data traffic flowing from enterprise applications are also converging onto packet networks. At most enterprises, Internet-based support for sales and marketing, for customer service and for supply chain management are now table stakes. Within the enterprise, the convergence of traditionally separate data and voice network infrastructure in the enterprise can reduce the cost of doing business.

For example, replacing legacy PBXs with VoIP is a popular converged application that can significantly reduce enterprise telecommunications costs. To prepare the infrastructure, local networks are upgraded and WAN links are re-evaluated. Legacy frame relay and private line networks can’t compete with low-cost, widely available Ethernet and IP-based services.

To serve the consumer and enterprise markets, carriers have traditionally built distinct networks for specific services--for example, a TDM network for voice and a packet based network for data. As demand grows for packet-based applications, and in order to reduce costs and complexity, carriers are working to eliminate distinct networks--with the goal of converging consumer and enterprise services onto a common optical transport architecture.

Sonet/SDH is the most commonly deployed metro aggregation transport architecture today, but it can be inefficient for transporting Ethernet, and it is limited to a maximum of 10Gbps. The market for Sonet/SDH has peaked, according to Infonetics Research, and they forecast it is beginning a long, steady decline. While the 400,000+ Sonet/SDH rings installed worldwide won’t disappear overnight, network operators face a common question: What optical network architecture is going to take my network through the next several decades?

Sonet was developed more than 20 years ago, when most packet protocols were still toddlers. While many large enterprises and carriers still use Sonet circuits to transport packet protocols, packet-based networking protocols such as IP and Ethernet have won the day because they lower the cost of building and operating networks.

Understandably, many legacy optical transport equipment providers have a vested interest in continuing the reign of Sonet/SDH and are presenting optical transport network (OTN) as the heir apparent. OTN provides considerable functionality (at a cost) to support a mix of legacy service interfaces--Sonet/SDH, ESCON and Fibre Channel--in addition to Ethernet. Unfortunately, OTN relies on a circuit-based optical network architecture just as Sonet does. With OTN, transponders are fixed in point-to-point paths and then a mix of circuit and packet services are groomed onto these fixed optical circuits. But bolting a packet service infrastructure onto a circuit-based optical network creates significant inefficiencies that dramatically increase total cost of ownership.

These inefficiencies arise in three areas: capital expenditures, optical span engineering and maintenance coordination. First, capital expenditures are high in circuit-based OTNs because transponders are provisioned as circuits, and bandwidth is stranded when traffic ebbs on a particular link. This wasted bandwidth reduces transponder utilization, requiring more transponders at a higher cost to carry the aggregate offered load of metro Ethernet traffic.

Second, optical span engineering in circuit-based OTNs requires forecasting traffic demand in point-to-point pairs, pre-provisioning fixed optical circuit paths between those points, and then calling the packet services team to let them know they can now provision packet services between those points. As traffic patterns change and new services are introduced, the OTN must be re-engineered, and optical circuits re-provisioned. This re-engineering and re-provisioning requires highly trained staff with complex IT tools, which are expenses that last long beyond the initial capital investment.

Finally, the delays resulting from OTN span engineering for transport necessarily delays the provisioning of packet services for revenue. Maintenance coordination between the packet service and circuit transport teams delays service turn-up and trouble resolution. Together, these costs can add up quickly and continue to grow over the life of a circuit-based OTN.

OTN or EOTN?
The network operators we are engaged with don’t want another Sonet--they want an Ethernet optical transport network (EOTN). Unlike an OTN with its circuit-based optical infrastructure, an EOTN delivers a packet-based optical infrastructure.

An EOTN enables any-to-any metro connectivity without the burden of pre-planning or provisioning of optical circuits. And as traffic patterns evolve, an EOTN automatically adapts, allocating bandwidth where it is needed without operator intervention. EOTNs change the fundamental economics of deploying optical networks by reducing both the capital and operational costs of building and maintaining a converged network.

The primary difference between an OTN and an EOTN is the degree of control plane interoperability between the Ethernet services layer and the optical transport network. Like Sonet, OTN provisions “dumb” point-to-point optical circuits. It’s referred to as dumb because each transponder deployed connects only two locations. Worse, because reconfiguring optical circuits requires operator intervention, OTN transfers the cost of adapting the transport network to changing packet service traffic patterns onto the network operator. These ongoing operational expenses result from the lack of control plane interoperability between the Ethernet services layer and the optical transport network.

Unlike OTN, EOTN links the Ethernet service control plane directly to the optical transport network, significantly lowering total cost of ownership. This Ethernet control plane interoperability enables the EOTN to switch any Ethernet packet to any wavelength, share transponder capacity among all network elements on the ring and dramatically increase transponder utilization. Increased transponder utilization reduces the number of transponders required to transport the aggregate offered load of metro Ethernet traffic, thus directly lowering capital expenditures. Operational expenditures are also lower with an EOTN because control plane interoperability minimizes or eliminates the expense of planning and pre-provisioning optical spans, both at initial installation and over the life of the network.

Two alternative architectures have emerged as contenders for EOTNs. The first relies on GMPLS control plane interoperability, and the second relies on Ethernet services control plane interoperability.

The first of these two approaches, the circuit EOTN, leverages the work of the Optical Internetworking Forum, which combines ITU ASON with IETF GMPLS to enable an MPLS data control plane to automatically reconfigure optical circuits to deliver packet services. This approach requires layering Ethernet switches over ROADMs that support OTN, then fusing the complex GMPLS optical transport control plane with the MPLS data service control plane for automated re-provisioning of optical circuits. In addition to the waste that results from deploying fixed optical circuits for packet services, this approach is expensive, complex and as yet unproven because it requires full interoperability of disparate network elements.

The second of these two approaches, the pure-packet EOTN, leverages the vast amount of work done by IEEE, IETF, MEF and ITU, which has elevated Ethernet to carrier grade. This second approach differs from the first in that the existing carrier Ethernet service control plane is used to configure the optical transport rather than GMPLS. And, this approach provides increased architectural flexibility, simplicity and lower cost of ownership when compared with the first.

Since protocols such as TCP/IP, MPLS, VPLS, QinQ, PBB, PBB-TE already use IEEE 802 MAC addressing to signal packet destination, and 802.1p fields to signal packet priority, a pure-packet EOTN will interoperate directly with any of these Layer 2 service architectures. Whereas the circuit EOTN approach assumes deployment of MPLS, the packet EOTN can work with any Ethernet-based service architecture--be it MPLS, VPLS, QinQ, PBB or any other protocol designed to run over a standard Layer 2 Ethernet.

This increased flexibility directly leads to lower costs of ownership. By eliminating the need for MPLS control plane interoperability, we eliminate the expense of implementing GMPLS in the optical transport and the requirement for MPLS at the service layer.

Newer technologies, such as optical burst switching, are now making it possible to build pure-packet EOTNs that obsolete the need for both optical circuits and GMPLS. Optical burst switching relies on proven Ethernet standards to control a packet-based photonic layer, so a transponder can be controlled via MPLS, VPLS, QinQ, PBB or PBB-TE to reconfigure transponders in real time on a packet-by-packet basis, thus increasing transponder efficiency and lowering capital and operational expenditures to deliver the lowest total cost of ownership.

Conclusion
As demand for optical networks rises in the metro for business and for consumer access aggregation, and as demand for Sonet/SDH declines, it is clear that optical networks are at a crossroads. The decision faced by network operators will come down to this – invest in legacy circuit-based optical transport networks, or invest in next generation packet-based optical transport networks.

Timon Sloane is vice president of marketing for Matisse Networks, www.matissenetworks.com. Sloane has more than 20 years of experience in data networking. He can be reached at (650) 938-5100.