Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Networks

MSE wins over edge routers

Posted: 01 Oct 2004 ?? ?Print Version ?Bookmark and Share

Keywords:edge router? mse? sla? l1? l3?

Edge routers are designed primarily for best-efforts-oriented traffic, as opposed to the real-time strict service-level agreement (SLA) requirements of the new services edge. By adapting edge routers to handle the additional services that providers are now required to support, architectural compromises are being made. These compromises affect all levels of the edge router, from internal software complexity to command line interface adaptation and more.

While edge routers were designed to transport IP packets efficiently, convergence has fundamentally changed the requirements for IP/MPLS edge products, which now must support scalable, multiservice, mission-critical traffic on a single product.

The demands of this emerging market require a new breed of device, built from the ground up to properly address L2 and L3 convergence at the multiservice edge (MSE). A properly architected MSE device offers distinct advantages over traditional edge routing technology in several key areas, such as service orientation, scalability, reliability, traffic management, QoS and ease of integration into existing L2 and L3 networks.

MSEs also help drive revenues by supporting a full suite of L2 and L3 services that can be easily delivered over a converged MPLS network, including a broad set of IP-based VPN, ATM, frame relay and Ethernet business connectivity services; broadband aggregation of the subscriber base; integration of wireless traffic on to the converged packet network; and IP services.

Also, MSEs must support a variety of access types, such as frame relay, ATM, Ethernet, packet-over-SONET/SDH (PoS) and DSL. Thus, it must also support a wide range of transmission rates, from sub-T1/E1 to OC-n and 10GbE. Support for multiple access media and service types opens up a range of possibilities for custom service definitions that can be supported from a single platform.

Improving scalability

The MSE architecture's flexibility and modularity allow I/O, data forwarding and control processing to be scaled independently. Thus, providers can tune their capital expenses for a tremendous range of point-of-presence requirements, delivering the improved "pay as you grow" capabilities required today while realizing the benefits of service offerings based on a single platform.

In typical edge router architectures, forwarding engines may be centralized or distributed, but both have their limitations. A centralized architecture has a single forwarding engine and I/O line cards. Forwarding capacity is capped by the single forwarding engine. Also, deep packet inspection negatively affects forwarding on all router interfaces.

A distributed router architecture provides an improvement by logically mating or integrating forwarding capability with I/O. If only a few ports are required, however, excess capacity on a line card may never be fully utilized and a larger inventory of spares may be required, increasing costs.

MSEs decouple I/O from forwarding engines, pairing them together logically. That lets providers add I/O and forwarding capacity independent of each other, as needs dictate.

For example, multiple I/O cards may be associated to a single forwarding engine. Decoupling of I/O from data forwarding also enables any service on any channel, for true multiservice environments. ATM, frame relay and PPP can be supported on any channel, without requiring a specific I/O line card per access technology. That reduces providers' guesswork for individual point-of-presence site requirements, while delivering line rate forwarding with flexibility and growth options.

As for control plane scalability, typical edge routers implement a single CPU for control-plane processing. For IP-VPN services, running multiple instances of routing protocols strains the single CPU-based control plane, gating the number of VPN instances.

MSEs provide multiprocessor control planes that deliver massive amounts of computing resources, letting providers gracefully scale their IP services as demanded. Processing and memory are simple field upgrades, easily accommodating service growth. MSEs ensure that control plane capacity does not limit new service growth, while conserving capital expenditures. Additionally, MSEs provide complete control and data plane separation.

Carrier-grade reliability

Where multiple services converge on a single device, any outage results in costly service interruptions. Therefore, new levels of reliability are unquestionably a prerequisite for multiservice network convergence.

Typical edge routers rely primarily on network-level protection mechanisms. They also feature some nodal-level protection mechanisms, such as nonstop forwarding and fixed hardware redundancy (graceful restart). At the service level, today's edge routers feature monolithic software architectures that provide insufficient protection between individual customer VPNs and routing protocol instances. The result is poor reliability if a failure occurs. A single customer failure may affect all customers on the system.

MSE reliability begins where edge routers leave off. MSEs deliver new levels of carrier-grade reliability at the network, node and service levels, reducing downtime from 150 mins per year to 5 mins.

At the network level, MSEs support fast convergence, fast reroute and MPLS global and local repair. At the nodal level, MSEs go beyond typical edge router hardware redundancy and SONET APS support by also providing hitless software migration, patching and rollback with zero downtime.

Also, MSEs support continuous routing in the event of a failure. "Nonstop routing" completely isolates a failure to the local MSE itself. A backup routing system takes over without adjacent routers' becoming aware that anything has gone wrong.

At the service level, MSEs provide superior protection and resource control. MSEs support software partitioning on a per-customer basis to isolate failures and ensure one service does not affect another service or the entire device. In so doing, each individual software processincluding VPN routing and forwarding entities for IP-VPN servicebecomes its own process, with tables for each customer and with no leaked routes between customers.

This ensures that if one customer encounters a fault, no others are affected.

Total system availability also depends on the reliability of operations, administration and maintenance (OA&M) systems. MSEs' OA&M systems provide reliable event collection and secure network operator authentication and authorization with customizable user privileges.

Traffic management

Since the converged services edge carries multiple applications, its traffic management architecture needs to reflect that fact by treating L2 and L3 traffic natively. Effective traffic management is critical to ensure customer SLAs are met, so it must be consistent across access types. Also, traffic management must ensure SLA consistency with the service set offered today, while enabling the required prioritization for new real-time, multimedia-rich services that have stringent network performance requirements.

Edge-router traffic management schemes are optimized for IP, with little concept of applications in a multiservice context. Some edge routers were first built as core routers and then redeployed to the edge. Hence, they were originally designed with a limited number of hardware queues at the port level, with no separation of queuing and forwarding class and no service awareness.

The MSE must deliver all services to customers with service awareness and quality treatment. Converging services require comprehensive and consistent traffic management/QoS systems at both L2 and L3. Effective traffic management is critical to ensure that customer SLAs are met, so it must be consistent across access types.

Today's typical edge-router traffic management schemes are optimized for IP, with very little concept of applications, particularly in a multiservice context. There are only a limited number of hardware queues, and no separation of queuing and forwarding class is possible.

MSEs use hierarchical scheduling to distinguish queuing and forwarding classes. This is essential, since such separation enables application-based queuing for extremely granular QoS/traffic prioritization. With application-based queuing, traffic can be weighted differently even if within the same forwarding class, thereby optimizing traffic management for true multiservice environments.

For example, best-effort IP-VPN traffic can be given higher priority than best-effort Internet access traffic, even though they are part of the same traffic class.

Intelligent integration

A design tenet of an MSE at the converged edge must be to easily integrate into existing networks to ensure smooth continuation of operations. That includes maintaining existing service definitions, whether they be at L3 for an IP-VPN or at L2 for an existing ATM service. It includes service interworking and mediation among ATM, FR and Ethernet access technologies, allowing end points from existing network devices to terminate on the new edge in a dynamic, resilient manner.

Finally, it includes network interworking with the ability to transport L2 networks over the MPLS infrastructure cost-effectively while maintaining QoS.

Edge routers were not designed to provide SLAs for existing L2 services. They were not designed to provide L2 signaling to establish connections dynamically, or for supporting the required QoS to transport L2 networks across an MPLS core.

Since MSEs were designed to support both L2 and L3 services, they can easily fit into existing networks and SLAs. MSEs support both L2 signaling and protocols. Thus they support dynamic and resilient service mediation, as well as network interworking, without requiring port-by-port connections. Whether in an IP network or a frame relay/ATM network, MSEs allow for a graceful migration and evolution path at a pace of a provider's choosing.

MSE platforms seamlessly interoperate with existing network elements, management systems and operations support systems while supporting current SLAs and service definitions. Integrating with existing L2 data networks is particularly important since it will enable the successful migration of business-grade data services over a converged infrastructure.

Ru Wadasinghe, Sr. Product Line Manager

John Beatty, Marketing Manager of Multiservice Provider Edge Division

Nortel Networks Ltd

Article Comments - MSE wins over edge routers
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top