Over the past several years, the industry has witnessed significant progress in Automatically Switched Optical Networks (ASON) and related optical control plane technology. Standards-based optical control plane technology provides functionality, reliability, and scalability for network operators, and includes the Internet Engineering Task Force’s Generalized Multi-Protocol Label Switching suite of protocols and the Optical Internetworking Forum’s UNI/NNI initiatives. This progress has resulted in a unified architecture that embraces the use of an Optical UNI to enable optimal IP/optical layer interworking.

As evidenced in Tier 1 wireline and wireless carrier networks around the world, there are significant economic benefits associated with the adoption of an ASON-empowered Optical Cross-Connect (OXC)-based transport infrastructure with mesh-based circuit routing and restoration. These providers derive economic benefit from the highly efficient use of backbone infrastructure and reduced recurring operational costs associated with network management and service provisioning.

IP-over-ASON has been proven to clearly show that multi-layer inter-working is ready for commercial deployment. In spite of this, it has been proposed that once routers with IP QoS and MPLS Fast Reroute (FRR) are deployed, there will be no need for an optical control plane; architectures that transport IP directly over DWDM are simpler and more cost effective. Although this approach sounds compelling in theory, it does not necessarily translate as such in practice.

IP-over-DWDM architecture suffers from inefficient utilization of DWDM infrastructure and router port capacity. Both inefficiencies are a consequence of routinely over-provisioning the IP layer to allow for link failures. IP-over-ASON solutions alleviate the inefficiency of straight IP by relegating the problems of link protection and capacity planning to the optical layer, while leaving - indeed, relying on - QoS management at the IP layer.

Based on network topology, network operators can realize economic gains of between 30% and 50%. Perhaps more importantly, these gains scale with increasing traffic demand, thereby enabling an IP-over-ASON service network with impressive robustness and longevity.

Limitations of IP onto Lambdas

The argument for the IP-over-DWDM approach typically centers on the reduction of CapEx and OpEx costs achieved through network simplification. In this model, IP routers, assisted by new protocols such as MPLS FFR, perform the switching, grooming, configuration, and restoration network functions. This model assumes that the only service provided by the optical layer is unprotected point-to-point trunks of fixed bandwidth.

In a world dominated by IP applications, the simplicity of an IP-over-DWDM architecture is appealing; however, it has several limitations that directly impact network scalability, efficiency, and reliability. Increasing levels of real-time voice and video-over-IP that require far more stringent QoS (latency, jitter, loss) schemes compared to the best-effort QoS of Internet surfing traffic exacerbate the challenge. Some of the limitations of this approach are described below:

· IP Only Service Quality: IP-over-DWDM may be an adequate transport backbone for IP traffic, but it does not effectively carry ATM/Frame, SONET/SDH/PDH or other traffic that requires grooming functionality. Although there is an ongoing effort to map everything onto IP via Pseudo Wire Emulation; this technology is immature for the backbone and introduces another set of OAM&P and interoperability issues. Empirical evidence of this limitation is seen as many operators would rather have their subscribers move to all IP, rather than accommodate these customers as-is on their ‘multi-service’ IP infrastructure.

· Scalability: Because of the non-deterministic nature of IP traffic, an IP-over-DWDM implementation requires over-allocation of capacity at each routing point to handle unpredictable transit traffic. For example, today the transport channel is a 10Gb/s wavelength. A 640 Gig router connected to six neighbors that each consumes 24 Gig of traffic will consume 18 x 10 Gig ports (3 x 10 Gig to 6 neighbors). After holding back half of the capacity for restoration & load balancing, this effectively leaves the router with 140 Gig of growth. Available capacity is even lower (perhaps at a deficit), when you consider that 50%-60% of router ports are used for interconnection within the POP. This introduces significant CapEx and OpEx burden in the form of port consumption and capacity ‘hold-back’ even before we address the inefficiency of tiered core IP networks – flatter is always better for IP service clouds.

· Layer 1 Failure Impact: 1 in 8 backbone failures are directly attributable to outages at the DWDM layer. The IP-over-DWDM approach requires the router to take care of all Layer 1 failures. Considering that a DWDM link cut can affect multiple 10 Gig channels, IP layer restoration could introduce a several-minute traffic hit for IGP/EGP routing convergence. To counter, operators face the choice of burdensome ongoing MPLS FRR planning or even more exaggerated over-provisioning. This renders the IP-over-DWDM solution too expensive.

Advantages of an IP-over-ASON Architecture

The alternative approach connects the core router to an ASON-empowered OXC before using the DWDM system (Figure 1). The ASON layer provides a set of functionalities that include provisioning and shared protection/restoration. Combined with efficient interfaces such as Ethernet (GigE and 10GigE) and the switching & grooming features in the OXC, this approach offers dynamic cost and operational benefits to the IP cloud.

Figure 1: ASON-Empowered Optical Network

ASON-empowered OXCs implemented at each backbone node provide optical control plane features at places with multiple fiber paths as well as each IP traffic collection and aggregation site. Through the use of this control plane, optical switch elements are intelligent and aware of their neighbors as well as the overall network topology. Changes in network state (e.g., DWDM outages, equipment failures, new service activation) are automatically communicated between elements using standard G.ASON protocols.

Traffic can be mapped across the transport facility using Ethernet or MPLS flow-based markers such as VLAN id or MPLS tag. This is in addition to the traditional physical port-based mappings provided by optical platforms. This leaves optical path topology and restoration to the ASON-empowered OXC and allows the router platforms to deal with traffic routing on Layer 2/3 terms - an environment clearly more comfortable to router vendors than the world of DWDM transport.

The IP-over-ASON network brings multiple provisioning, protection and restoration schemes to the network with the control plane to manage their operations. This includes the ability to deploy traditional rings, as well as sub-50ms end-to-end path protection and dynamic mesh restoration. These can be used in a hierarchical fashion to provide a robust network able to set up new service dynamically as well as recover from multiple diverse outage situations.

The IP-over-ASON advantages include:

· Topology Independence: IP network routing topology is not affected as links are established or removed. Additional capacity can directly join existing link(s) through the router's link aggregation scheme. Thus, concerns that topology changes in an IP-over-ASON architecture result in slow routing convergence are a non-factor.

· Superior Transport Capacity Utilization: Each backbone link's utilization ratio can improve significantly as dynamically established links can be used to carry traffic for multiple destinations, based on the ASON OXC's efficient traffic grooming capabilities (e.g. Ethernet or MPLS flow). With this scheme, multiple routers can be aggregated through the same backbone transport link. This flexibility cannot be achieved by IP-over-DWDM because, in that architecture, each router port has to be hardwired to a specific fiber destination.

· Increase in Effective Router Capacity & Service: As stated earlier (‘flatter is always better’), backbone routers typically use interconnect links to form a fully meshed topology. Once a router is down, the connection to it will not be useful anyway. However, if this port interconnection is made through the adjacent ASON OXC, the same intra-POP link can be reused to establish a new connection.

Additionally, this optimization capability further reduces the number of expensive router ports required in an IP-over-ASON implementation. The number of connections required by the routers connecting to the OXC can start from a very small number (i.e., one) and grow incrementally, based on increased total capacity demands on that router. Since a single router interface is able to support many virtual sessions between multiple peer routers, the operator is able achieve very high utilizations from these service platforms – without consuming ports just to establish the right topology.

Conclusion
In comparison to the IP-over-DWDM baseline model, the IP-over-ASON network typically introduces between 30% and 50% CapEx savings. The savings are achieved both by improvements to DWDM transport efficiency as well as substantially increased backbone router efficiency. This reduces the number of expensive router ports needed to support traffic demands. Additionally, dynamic protection schemes also eliminate some of the latent protection capacity required.

Although QoS-aware routers have been widely deployed, most carriers are still examining the over-provisioned IP-over-DWDM scheme. Repeatedly, carriers have realized that compared with standard IP-over-DWDM architecture, the IP-over-ASON networks offer compelling benefits for service providers, including:

· Dynamic provisioning, addressing non-deterministic, unpredictable IP traffic
· Significant total network CapEx reduction
· Enhanced network protection against various network failures
· Facilitation of new IP services by using Layer 2/3 flow-based grooming onto transport

Deploying the right transport architecture based on ASON empowered Optical Cross Connects unlocks the QoS capabilities of today’s core switch/routers while reducing network operations CapEx and OpEx for the long term.

Robert Adams is Director of Product Marketing for Sycamore Networks.