Connections +

Convergence at the network edge

The drive to simplify new service deployments compels service providers to converge Layer 2 and Layer 3 networks and services into a new multiservice edge. The problem is existing edge devices are not up to the challenge.

September 1, 2004  

Print this page

Today, using wireless and other broadband technologies, handheld devices smaller than a wallet enable multimedia communication and collaboration from any location by transmitting volumes of digital voice, video, and data traffic.

PCs, workstations, and servers are linked into high-speed virtual networks that span cities, countries, and continents and share gigabits of information.

Customers can receive differentiated, tailored, private services that securely traverse shared public networks of hybrid media.

Now that Utopian visions of anywhere-anytime communication are coming to fruition, why aren’t service providers wallowing in profits?

Consumers and enterprises are enjoying halcyon days of choice and value-a veritable buffet of multimedia services at attractive rates, bundled to please, on demand.

If service providers are so successful providing what their customers want and need — and they are — why are bottom lines so bleak for so many?

Revenues from voice traffic are steadily on the decline, and traditional data services such as Frame Relay and ATM are already starting to plateau.

These trends are partly spurred by the burgeoning popularity of IP-based voice and data services, but profits are not keeping pace. Not even close.

Even though IP traffic has doubled each year for the past three years, revenues from IP services have only risen 15% in that time, according to a report released last year from the Yankee Group. So what gives?

For one, users of the Internet and IP-based intranets demand plentiful and ever-increasing bandwidth but are reluctant to pay much for it.

For example, a typical residential customer would willingly pay seven cents a minute for long-distance voice calls, yet bristle at anything more than $35 per month for unlimited high-speed Internet access that is almost two orders of magnitude cheaper — at roughly $0.0008 per minute.

Furthermore, business and residential customers expect a smorgasbord of features and services that were unimaginable a decade ago.

Service providers have responded admirably by expanding access options to include everything from traditional telephony to wireless and cellular to cable and satellite and DSL, in addition to the usual array of dedicated data lines.

In the core, they’ve deployed IP for IP services, such as IP-VPNs, and ATM for Layer 2 data and packetized voice traffic.

Behind the scenes, they’ve deployed gateways, servers, and switches that empower the IP network for “high-touch” services such as stateful firewalls, network address translation (NAT), and intrusion detection. (see Figure 1)

Unfortunately, satisfying customer expectations has resulted in less-than-satisfying architectures — typically using multiple networks to deliver the full range of required services. This complexity represents a capex/opex headache.

As revenues for voice and data services flatten or decline, the care and feeding of these multiple networks is becoming unwieldy and inefficient. Yet for all the cost of supporting multiple infrastructures, this equipment still has years of useful, revenue-generating life.

A fuzzy reality

There’s rarely a business case for scrapping what is already in place and working well-even if a more efficient alternative could be bought anew.

What is the answer for forward-thinking providers who want to extract maximum value from existing network investments, while recognizing that costs are escalating faster than revenues for even the most coveted services?

The fuzzy (but impending) reality of convergence: Bearing all business realities and technology advances in mind, the time is fast approaching for a new network architecture that converges Layer 2 and Layer 3 services onto a single edge for transport over a unified core.

“Convergence” and “multiservice” as network concepts have been talked about for so long and in fact, vendors have been evolving their network devices toward the multiservice ideal for some time.

Traditionally, “multiservice” in data networks has meant adding multiple services from the same layer of the Open Systems Interconnection (OSI) model-for example, supporting both Frame Relay and ATM on the same device.

Today, Multiprotocol Label Switching (MPLS) is emerging as the de facto standard to enable convergence of networks and services, by being media-agnostic, offering a standardized encapsulation technique, and bringing deterministic QoS and traffic management to IP.

This reality is coming closer as key players — research firms, vendors, and major carriers (who are in the process of deploying unified IP/MPLS core networks) — come to agreement on IP/MPLS as the de facto standard for this edge convergence.

The Multiservice Edge (MSE) is a new market segment of devices focused specifically on the convergence of Layer 2 and Layer 3 data services at the edge of an IP/MPLS network.

MSE platforms leverage emerging standards to combine select functionality of edge routers, multiservice WAN switches, Ethernet switches, and IP services switches into a single edge device.

This broad range of services is essential to support today’s revenue streams, while providing a seamless migration path to new services over a cost-reduced infrastructure.

Nortel Networks believes there are five key attributes that are essential for a new class of an MSE network device.

In order to redefine the concept of “multiservice” and in turn, redefine the ROI of multiservice networks, this new network element must be able to:

1. Support traditional and emerging Layer 2/3 services: In order to streamline the network while supporting a broad range of customer service requirements, the ideal multiservice edge platform must be, well, multiservice.

This new definition of multiservice means it must support a broad range of Layer 2 (Ethernet, virtual private LAN service, Frame Relay, and ATM) and Layer 3 (IP-VPN, broadband Internet access) services and also maximize revenue potential with value-added services, such as NAT, firewalls, and intrusion detection security services.

It must support a variety of access types, such as Frame Relay, ATM, Ethernet, Packet over SONET/SDH (POS), and DSL. By default, that means it also must support a wide range of transmission rates, from sub-T1/E1 to OC-n and 10 Gigabit Ethernet..

2. Provide true carrier-grade reliability at multiple levels: As the industry moves toward a converged infrastructure, IP networks will be entrusted with mission-critical voice, video, and data traffic-and a single device could support a large number of subscribers.

For any service level agreement (SLA) to be meaningful, the multiservice edge platform must be able to uphold stringent, carrier-grade standards.

That kind of reliability can only be achieved when it is integral to the design at multiple levels — node, network, and service.

At the node level, it involves redundant common equipment with flexible sparing capabilities and a hardened operating system.

On the networking level, it should include non-stop routing and intelligent, resilient re-routing, as well as hitless equipment switchovers and software upgrades. At the service level, it addresses such features as journaling (maintaining connection-state awareness across equipment switchovers/failures), virtualized software processes, and software process sparing.

3. Provide a modular, flexible hardware and software architecture: Service providers can no longer afford to deploy a daisy-chained stack of small, lower-speed devices to support growth. Nor does today’s lean profit model justify wholesale change-out of equipment to support new services.

The ideal multiservice edge platform must have a modular hardware and software design that can be cost-effectively deployed for a single service offering, yet be able to evolve on a “pay as you grow” basis to support a complete suite of new and existing Layer 2 and Layer 3 services.

The design must have the physical flexibility to cost-effectively scale up, to meet the requirements of large central off
ices (COs), or down, to be deployed in small or medium COs. The device should have an independently scalable, distributed architecture for input/output, control plane, and data plane processing — allowing massive subscriber and services growth by adding capacity to existing elements only where it is required.

4. Optimize traffic management for the multiservice environment: Imagine if the highway system had parallel but separate lanes for each type of traffic. Commuters in private cars drove in one lane, SUVs and pickup trucks in another lane, carpoolers in a third lane, and taxis and buses in yet another.

It sounds costly and inefficient, and it is. But this is essentially the way traffic management is handled in today’s multi-network architectures due to historical build-outs and the technology available at the time.

Different services have their unique traffic management schemes, with each optimized for utilization and performance within that service but not across different services or media.

Hence, when transporting one service overlaid on top of another, for example deploying a Layer 3 service over a Layer 2 transport technology, optimal treatment of the originating service traffic (Layer 3 packets in this example) is not possible since it is subject to a different (Layer 2) traffic management scheme.

As the lines between different services start to blur, service providers need to be able to offer a consistent SLA based on the service, independent of the media.

To accomplish this, the multiservice edge device must have an integrated Layer 2/3 traffic management scheme that is applied to the end-service and/or customer, while remaining consistent across all access and transport media.

5. Integrate intelligently into the existing infrastructure: The existing network, with its multiple layers of service-specific components and protocols, may be complex, but it’s still worthy and has useful life. It’s widely deployed, it’s generating revenue, and it may not be fully depreciated. You should be able to extract maximum value out of that existing asset.

That means the new multiservice network edge platform must seamlessly inter-work with existing network elements, management systems and operations support systems, while supporting current SLAs and service definitions.

Integrating with existing Layer 2 data networks is particularly important, since it will enable the successful migration of business-grade data services over a converged IP/MPLS infrastructure

This strategy wrings maximum value out of existing infrastructure — a critical boon for capital expense — while trimming operating expense through reduced requirements for training and deployment, and allowing providers to migrate to MSE devices on their own terms and timelines.

A multiservice edge device that satisfies these five key attributes will be an asset for any service provider’s convergence strategy, whether the foremost goal is to enhance service delivery and revenues, evolve to a more cost-effective infrastructure, minimize costs or achieve all three.

Chris Chartrand is Senior Manager, Multiservice Edge Marketing at Nortel Networks and has spent the last six years working in various roles in the company’s data networking group. He can be reached at