With the industrywide rollout of 100-Gb/s coherent wavelength technologies and 500-Gb/s superchannels underway for addressing scalable fiber capacity, service providers are increasingly looking to deploy DWDM transport systems with integrated OTN switching capabilities over increasingly meshed transport topologies. These converged platforms, combined with an intelligent GMPLS control plane, enable rapid provisioning (and de-provisioning) of sub-wavelength bandwidth services at granularities from 1 Gb/s to 100 Gb/s, using any wavelength resource, and delivering the service between any two network endpoints, along with providing valuable grooming and digital protection features.
This essentially provides a new virtualization capability of the optical resources and enables operators to decouple the bandwidth services (Ethernet/OTN/MPLS) from the shared pool of underlying optical resources comprising the transmission layer. This is in contrast to the operationally challenged traditional transponder/muxponder model, where bandwidth services are statically mapped into wavelengths that are switched by reconfigurable optical add/drop multiplexers (ROADMs) and delivered to two fixed network endpoints.
The Role of Intelligent Transport in Multi-Layer Networks
The emergence of the converged transport system and its inherent bandwidth virtualization capability enables a new level of agility at the transport layer that could have a significant impact on how multi-layer core networks are built, especially when software-defined networking (SDN) concepts, such as a logically centralized network view, are leveraged. Instead of the conventional approach of performing all bandwidth management functions and recovery within the IP/MPLS router layer and using optical wavelengths solely as dumb point-to-point delivery pipes between routers, the emerging intelligent transport network offers a lower-cost approach. By leveraging converged DWDM and OTN switching for provisioning flexible, mesh-based transport services and employing new technologies such as sub-50ms shared mesh protection within the transport layer, the intelligent transport network can provide a lower-cost alternative for implementing next-generation networks by reducing the total traffic processing required within the router layer.
In order to take full advantage of this intelligent transport network, service providers need to consider more streamlined methods for integrating and coordinating the operations at both the router layer and the optical layer.
Current practices are typically done on a per-network-layer basis, with planning, provisioning, and resource optimization performed solely within each layer, and with little or no interlayer topology information sharing. Additionally, bandwidth provided by the optical layer is often manually deployed as a transponder or muxponder and treated as a static resource by the router layer, as opposed to a dynamic, on-demand resource that can behave more like a utility, like electricity.
In this type of conventional core-network architecture, routers are interconnected via a transport network and typically only possess knowledge of the IP-layer topology, not the network layer closest to the physical fiber infrastructure itself — i.e., the transport layer. Links interconnecting routers into a partial mesh topology that may appear to be diverse may in reality be delivered over shared transport systems that might fail or physical facilities (e.g., fiber pairs or conduits) that might get cut, creating multi-link failures at the IP layer that can significantly disrupt services.
Furthermore, because of the connectionless nature of routers, and the topologies deployed today, routed networks often rely upon intermediate core routers to forward packets on to their next-hop neighbor. While the statistics will vary from network to network, a significant amount of the traffic actually transits through a router, destined for another router along the packet’s data path. In this approach, if a congested router link is in need of more bandwidth (perhaps as a consequence of unnecessarily carrying transit traffic) then a request is typically submitted to the network operator’s transport organization for incremental wavelengths, which today often involves manual processes and procedures that can take days or weeks.
SDN for Multi-Layer Provisioning & Optimization
Consider now a paradigm where transport bandwidth between routers transforms into an on-demand, variably sized utility, drawn from a pool of capacity that is sharable across the network instead of a static dedicated resource. What if the IP topology could adapt more dynamically to meet the spontaneous needs of the applications layer, and traffic could be transported more efficiently through the multi-layer network? By extending SDN to the intelligent transport network, a centralized view of topology and resources at both the IP and converged transport layers can be attained, and through this global view, traffic and resource management across networking layers can be orchestrated, provisioned, and automated. This level of coordination and automation across network layers not only presents huge operational savings but can also dramatically speed bandwidth service delivery, enable application driven networking, and improve the carrier’s competitiveness.
Additionally, the SDN control layer can house new software intelligence that can analyze this pool of virtualized resources across all layers and calculate optimal paths based on operator-specified metrics (cost, latency, power usage, etc.) and service requirements. It can also conceivably take into consideration other important data beyond the current network state, such as future traffic demands or network modifications. For example, the most cost-effective approach for a set of large-elephant flows with stringent real-time SLA requirements may be to keep the traffic at the transport layer using a dedicated horizontal slice of the transport layer (e.g, an optical VPN). This not only gives the dedicated transport bandwidth needed by the specific traffic type, but can also help prevent this traffic from 1) consuming router resources for transit functions and 2) impacting the smaller or shorter-lived “mouse” flows.
In a different traffic scenario, the optimal solution might involve creating an overlay multi-layer VPN that keeps the flows primarily in the transport layer but employs a router’s capabilities at just a few strategic points, as opposed to unnecessarily touching each router along the traffic flow’s path. This concept of router bypass isn’t new but, when employed, has traditionally been determined using offline optimization methods and manual optical bypass methods. With an SDN approach, reactive real-time analytics and automated optimization could dramatically enhance the efficiency of the routed network, thus lowering both router capex and energy consumption.
Optical Capacity On-Demand — Dream or Reality?
On-demand delivery of digital transport bandwidth can easily be supported in mesh networks that leverage optical transport solutions with integrated digital switching, but only up to the amount of optical capacity deployed. One of the challenges operators face is keeping ahead of the demand curve, and ensuring sufficient capacity is available in the network (without overdeploying capex) while also ensuring their processes and supply chains are tuned to engineer, deliver, install, and turn-up transponders/muxponders as quickly as possible. Depending on the rate bandwidth is consumed and the variability in time, this can prove to be a daunting task. Increases in traffic variability, transience, and bandwidth-on-demand applications can exacerbate the issue.
For transport bandwidth-on-demand to work, the network requires spare optical capacity that can serve as a buffer. With a transponder-based network architecture, this means pre-deploying spare optical transponders/muxponders — a financially challenging proposition for most operators. Economics aside, it would require some automated means for steering, configuring, and turning up the wavelengths.
While SDN can’t aid in solving this capex challenge, there is some discussion around leveraging SDN to program wavelengths. For core long-haul transmission, this would require complex controls, as solutions for optimizing analog optical parameters and maximizing both performance and capacity vary amongst vendors and are not well suited for conforming to a common abstraction. As Glenn Wellbrock from Verizon stated at the 2013 Packet Optical Transport Evolution event hosted by Light Reading, “Lots of physics is used to enhance transmission performance. [An] SDN controller will not be able to manage this constant evolution.”
A better approach to realizing this vision is to leverage larger pools of deployable optical capacity such as pre-lit 500 Gb/s or 1 Tb/s superchannels that can be activated and paid for in 100-Gb/s increments. For example, imagine a 500 Gb/s super-channel pre-lit between two cities but with only 100 Gb/s activated and paid for initially. As part of enabling a bandwidth-on-demand solution, dynamic activation of the second, third, fourth, and fifth incremental 100 Gb/s chunks of optical capacity can certainly be done via software licensing and is easily controllable by SDN.
There is significant value to be gained from extending SDN to the emerging converged transport layer — optical VPNs, multi-layer provisioning, and multi-layer optimization, to name a few. While the newfound flexibility and agility in the intelligent transport layer can alone significantly reshape how multi-layer networks are architected, a centralized network topology and intelligent orchestration and provisioning functions are essential elements in making these multi-layer networks operational and minimizing manual procedures. With the promise of multi-layer SDN, these new networking applications can help reduce operational complexity, reduce capital costs, improve overall network resource utilization, lower power consumption and speed service delivery.