Bringing intelligence to the optical edge

June 1, 2001
Metro Networks

Shifting intelligence from the core to the edge of the network will simplify multiple services and deal with rising traffic volume.

GERRY WHITE, RiverDelta Networks

Fiber-optic technology and applications are creating tremendous opportunities for delivering new services, but to take advantage of the enhanced network, service providers must address the challenge of delivering end-to-end quality-of-service (QoS) guarantees across access, metropolitan, and core networks. Providers must police, isolate, and groom traffic at the edge of the network to enable tiered services across the optical infrastructure.

The volume of data traffic continues to increase at an exponential rate and the character of the traffic itself is evolving from best-effort data transmission into the delivery of multiple services-each with its own QoS requirements. To transport this growing volume of traffic, carriers are aggressively building out the optical infrastructure. To simplify and scale the delivery of multiple services and cope with constantly increasing traffic volume, the intelligence in the network must shift from the core to the edge.

This shift requires the deployment of high-performance optical edge routers at the edge of the broadband access network. Whether the broadband access network is cable, DSL, satellite, mobile wireless, fixed wireless, or optical, the edge between the access network and the optical metropolitan-area network (MAN) requires advanced intelligence so that sophisticated traffic grooming and forwarding can take place in a distributed fashion at the periphery of the network.

This approach has already been proven by the evolution of Internet QoS standards, which simplify processing in the core network by having packets classified and groomed by the edge devices. Diverse traffic flows have varying QoS requirements, and carriers require intelligent edge routing so traffic flows can be classified and treated before entering the optical metro or core network. By providing increased intelligence at the edge of the broadband network, telecommunications providers will be able to ensure the successful delivery of end-to-end QoS across both the broadband access network and core optical infrastructure.
Figure 1. Current architectures are based on interconnecting multiple SONET rings onto higher-speed metropolitan topologies for transport to core backbone networks.

Optical amplification and transmission techniques eliminate the need for electrical regeneration, even in long-haul backbones. Increasing channel capacity and rapidly falling costs per channel make optical transport networks both economically viable and attractive. Carriers can connect broadband access networks onto optical core networks through SONET metro networks, or they can further extend the optical network by deploying optical switches connected to broadband access networks.

Many current carrier networks for Internet Protocol (IP) traffic consist of an underlying SONET transport network with a hierarchical IP routing overlay. The transport network is typically built from tiers of SONET rings (see Figure 1), providing a time-division multiplexing (TDM) infrastructure well suited for aggregating voice transmissions with predictable traffic patterns and well-defined data rates-but it's much less efficient for bursty traffic.
Figure 2. Hierarchical routed core and access networks will evolve toward flatter optical networks in the future.

The IP infrastructure is currently built with a hierarchical overlay infrastructure of routers (see Figure 2). Various routing protocols are used to establish a mesh of logical, high-speed, point-to-point connections between routers over the underlying infrastructure. End-to-end IP connectivity is achieved through multihop packet forwarding along the paths between routers. Management and control is segmented at different layers of the hierarchy and different administrative domains within these layers.

Optical carriers can take advantage of the latest technical developments to overcome the limitations of current networks. A next-generation architecture is required, and it will be based on deploying intelligence at the edge with a less intelligent high-speed optical switched core. Each traffic flow can be classified at the edge of the optical network and assigned a different lightwave-or "lambda"-for swift and efficient routing to its end destination.

The core is a lambda-based transport network constructed as a mesh of optical paths with IP routing used to select the best path between edge routers. The optical path comprises a combination of lambdas and physical fiber-optic cables. This design eliminates the need for optical-to-electrical conversion or signal regeneration. Operators can either use add/drop multiplexers (ADMs) to provide access to SONET rings or deploy optical switches in access locations for direct access to a flat optical network. This approach eliminates SONET ADM and IP routing/switching overlays within the core and results in an IP service delivery infrastructure with a single-level, single-hop, logical mesh interconnection of IP routers. Control of the optical paths is analogous to Multiprotocol Label Switching (MPLS) path control and can use similar mechanisms.

The edge routers and optical mesh infrastructure work in concert to achieve end-to-end service flow definition and management. The QoS levels are defined for given paths across the optical core, and traffic engineering is performed using lambda switched paths (LSPs) in the same manner as is done for MPLS. The edge routers perform per-service flow policing and forwarding and map traffic to the appropriate LSPs based on addressing information, QoS requirements, and routing policies.
Figure 3. High-capacity optical networks are being created to aggregate traffic from multiple access networks.

This approach allows the core optical network to be protocol-independent, with the edge devices providing translation and encapsulation as required for the transport of IP, ATM, MPLS, and native video streams. The architecture does not preclude a ring topology in the core, but a mesh topology is expected to be dominant since it offers cost advantages (redundant paths are 1:N rather than 1:1).

Mesh topologies are also easier to upgrade since these upgrades can be performed link-by-link without the need to upgrade an entire ring. Flat optical core networks with intelligence at the edge are being deployed to accommodate multiple access technologies. These next-generation optical networks will allow carriers to create a high-capacity core network that can be fed from cable infrastructure, DSL, satellite, mobile wireless, fixed wireless, or optical access networks (see Figure 3). All these technologies are coming together onto the optical core to enable the performance and capacity required by the explosive demand for bandwidth and services.

The core network becomes protocol-agnostic, and lambda switching allows high-performance delivery of diverse services, applications, and content end-to-end from various access networks to the core. Local-access networks can be connected to the core network through the optical edge routers using conventional IP routing, ATM, or MPLS. If traffic between the two endpoints does not justify a complete lambda, then MPLS tunnels may be used as an additional multiplexing layer to create an access hierarchy.

Increased intelligence at the network edge allows carriers to build protocol-agnostic optical core networks that implement a relatively limited set of functions. The primary function of the optical core, therefore, is to provide high-speed optical transport of one or more wavelengths between edge routers. To achieve that, the core must also maintain a mechanism to manage paths between the edge routers. To support the high levels of availability required, mechanisms for fault detection and path restoration are required.

Alternate paths can be set up ahead of time to allow for fast path switching following the failure of any given paths. These paths will be maintained as multiple routes to a destination by the IP routing protocols, with the lowest-cost path normally active. On detection of a path failure, traffic can be diverted to the alternate route immediately by the edge router without needing to wait for network-wide route convergence. If faster restoration is required, it can be provided by physical restoration techniques at the optical layer.

In the next-generation optical network, edge routers will require the intelligence to classify packets and police traffic flows in real time. That requires high-performance packet processing to support classification and forwarding of hundreds-even thousands-of traffic types. High-density, carrier-class edge routers are required so traffic flows from access networks can be efficiently classified and routed, either onto SONET rings or directly onto the optical core via MPLS LSPs, ATM virtual circuits (VCs), or dedicated lambdas. Decoupling the intelligent edge from the core allows the core and access networks to evolve semi-independently as new technologies are introduced. With carrier-class edge routing, operators can deploy increased intelligence at the edge of the hybrid fiber/coaxial (HFC) network to make the edge network core-agnostic. The core may be based on hierarchical routing or on a flat optical network, and operators can evolve the core network as newer technologies are introduced.

The optical edge router must act as an aggregation point for IP traffic from the access network. In the access network, QoS must be applied on a per-flow basis; in the core network, aggregated QoS will be used. Therefore, the edge router must map individual flows into macro flows to which backbone QoS treatments can be applied. A macro flow, consisting of multiple individual flows with common characteristics, is mapped to a traffic-engineered LSP for the appropriate destination edge router.
Figure 4. High-performance intelligent edge routers must aggregate flows and convert them into logical bundles for transmission over wavelengths through the optical core.

That allows multiple access networks to use a common core mechanism. Traffic received from the core network must be mapped into individual flows for transmission to the access network with the required QoS. That requires the edge router to be capable of classifying traffic at the line rate of the core interfaces, which demands hardware-based, carrier-class performance levels (see Figure 4).

The metro or core network may be operated by the broadband access provider or by carriers offering wholesale optical transport to multiple broadband access providers. The edge routers must be able to map traffic flows from multiple access networks across both the MAN and across the core backbone networks of these providers. That allows carriers to provide end-to-end QoS treatments across their core networks and across the backbone networks of their revenue-sharing partners.

Next-generation routers must deliver wire-speed flow classification, policing, and QoS assignment and support hierarchical, flow-based queuing. They must also have high-capacity WAN/MAN interfaces and offer carrier-class implementations of major routing protocols such as Border Gateway Protocol (BGP-4), open shortest path first, intermediate system to intermediate system, and MPLS. Edge routers must offer per-flow accounting, configuration, and management and enable end-to-end service creation and per-flow statistics for billing, accounting, and reconciliation.

These routers must offer the performance and functionality necessary for aggregating many flows onto the optical core network. Older router platforms lack the performance, scalability, and flexibility needed to support demanding requirements at the edge of the optical network. Intelligent routers at the edge of the network allow telecom providers to consolidate broadband access traffic flows and efficiently route them across the optical network. Carriers can rapidly introduce data, voice, and multimedia services for both corporate and residential subscribers and deliver QoS levels end-to-end across both the access and optical core networks.

Optical-network operators need to en sure the edge routers can scale to support new services and new business partnerships with broadband access providers. This approach requires carrier-class routing software implementations that can scale as operators expand their networks and develop new partnerships with partners. Carrier-grade routing software can ensure high availability for the network and superior network reliability, while providing increased intelligence at the edge of the cable and optical networks.

Carriers cannot afford to experience network downtime, so edge routers need to offer carrier-class reliability and high levels of hardware redundancy to ensure 99.999% availability. As optical providers continue to offer services that demand ultimate reliability, such as toll-quality voice, they cannot allow the edge router to impact availability. Edge routers, therefore, need to support all the traditional central-office operational requirements such as minimal disruptive software upgrades, redundancy, live insertion, version roll back, integration into the alarming scheme, and full Network Equipment Building Standards (NEBS) compliance. Any failure should result in swift service restoration through the activation of alternate hardware components, network links, and route paths.

The edge router must be able to classify data from the access network and map between the aggregated traffic engineering mechanisms of the core and the per-flow QoS flows of the access network (see Figure 5). Optical backbone providers may select between various mechanisms such as MPLS, ATM VCs, and the differentiated services (DiffServ) protocol for guaranteeing QoS through the core network. By identifying and classifying individual traffic flows at wire speed at the edge of the network, the router can map the access flows to the desired protocols used on the optical backbone. Operators can, therefore, support advanced service-level agreement (SLA) parameters such as maximum bandwidth allocation, minimum bandwidth guarantees, bounded delays, and bounded jitter.
Figure 5. This example shows how an optical provider can use intelligent edge routers with high-performance interfaces to allow a cable operator to deliver enhanced services through the metro network to provide intelligent transport to multiple back-end providers.

They can identify QoS parameters both statically (such as old/silver/bronze services) and dynamically (for services such as voice call setup). The edge router, therefore, has to interact with the QoS management systems of both the access network and optical core network to request and release resources such as light paths, Data-Over-Cable Service Interface Specification (DOCSIS) service interface definitions (SIDs), or MPLS LSPs.

Per-flow queuing enables the router to isolate the traffic of different services and different providers at the edge of the access network. Intelligent edge routers designed for per-flow policing and traffic shaping at wire speed allow providers to deliver SLAs on a short-term or long-term basis. With high-performance, content-aware packet classification at wire speed, providers can offer customized QoS levels and guaranteed SLAs. Optical providers can therefore isolate each of its wholesale access partners so none of them can impact other wholesale customers. That way, any overload or misbehavior within a partner's service can be isolated for that particular service.

Legacy routers lack the performance, scalability, and per-flow processing necessary to implement QoS effectively. They cannot perform source-based, content-aware routing because the per-flow packet classification and QoS required is beyond their processing capacity. But with today's high-powered silicon and advances in QoS theory, next-generation edge routers can examine individual traffic flows at wire speed. Thus, the stage is set for generating incremental wholesale revenues in a multiprovider environment.

Operators and their partners can implement policy-based routing in a multiprovider network environment. Intelligent edge routers can route IP packets based on any fields or set of fields in the IP header in real time. Source address routing is the simplest form of policy-based routing where traffic for a given service provider is recognized by the IP address and routed appropriately across a multiprovider network.

The routing may be determined by looking at the source IP address to determine to which service-provider partner the IP address belongs, then routing the traffic to that partner for handling. Such examination allows the router to implement more sophisticated QoS policies than those possible by simply looking at the data's destination address.

Unlike tunneling approaches, policy-based routing is designed for broadband IP infrastructure. Because the router can look at the individual application flows, it can extend the QoS capabilities of the access network. It can assign QoS and routing policies based on parameters such as service-provider, subscriber, and application. Because policy-based routing conforms to IP standards, current features such as transparent IP multicast can be supported. Policy-based routing can also take advantage of future developments in IP standards, such as the rollout of MPLS. Keeping up with the data rate of the optical interfaces requires hardware-based processing and wire-speed performance.

Optical networks need added intelligence and performance at the network edge to efficiently link broadband access infrastructure to the optical MAN or core backbone networks of multiple pro viders. Intelligent edge-routing platforms can deliver the carrier-class performance and reliability essential to the delivery of next-generation services requiring end-to-end QoS in a multiprovider network.

Standards-based policy routing, combined with full-featured, scalable routing protocol support, enables interoperability with equipment from multiple vendors. Thus, MAN operators can deliver transport services over optical networks to multiple access providers with diverse network topologies. High levels of reliability are needed at the network edge so broadband providers can reliably deploy increased intelligence at the edge of the network to support critical applications and services.

These are exciting times for optical-network providers, as they exploit market demand and emerging technologies to create multiple revenue streams and build closer bonds with residential and corporate subscribers. Optical networking is enabling the delivery of robust new services that traverse the broadband access network and allow subscribers to transparently select from a broad range of services and providers.

Operators are evolving from hierarchical routed networks to flat networks so they can leverage the capacity, performance, and efficiency of optical networking-and as they deploy increased intelligence at the network edge, they can gain major business advantages. Next-generation optical edge routers allow providers to maximize billable revenue over core and metro networks, and increased intelligence at the edge of the network is allowing access networks to benefit from the capacity and efficiency of optical backbones.

Gerry White is vice president and chief technical officer for RiverDelta Networks (Tewksbury, MA). He can be reached at the company's Website,