It’s all over but the shouting. Ethernet access is the way to go. Service providers stand to revolutionize their business models and total cost of infrastructure ownership by replacing aging and disparate voice, data and video connections with one fast, flexible and inexpensive pipe.

Hold on, though. There are a few hitches. Customers must not be asked to make radical changes on the service provider’s timetable. Overnight replacement of reliable legacy equipment is not an option for many enterprise customers. Nor will many give up the mission-critical data and communications delivery guarantees they now enjoy. Ethernet access technology alone must rise to challenges on both fronts.

No matter how low the price, sacrificing quality would end up costing the customer more. Imagine an accountant that offers service tiers with graduated pricing: Gold-level service with a three-percent error guarantee; silver, 15 percent; and bronze, 25 percent, all with a limited liability should an audit ensue. Even the gold service--no matter how fast or convenient--would have to be priced near zero to lure IRS-sensitive customers away from the tried-and-true, co-signed liability guarantee.

Telecommunications customers used to the guaranteed quality of time-tested TDM voice, frame relay and ATM services could face a similar choice as their service providers’ rush to deploy fast, convenient and lower-cost Ethernet and Internet protocol (IP) technologies.

In short, successful Ethernet services must not only cut costs and increase service value, they must also simultaneously match the quality-of-service (QoS) guarantees of proven wide area network (WAN) access and backbone services. Granular, deterministic QoS provides the key not only to service level agreement (SLA) documentation, but also to the development of premium services that differentiate the provider and generate revenue growth.

Standard Ethernet's shortcomings are well known. Even with the aid of IP/multi-protocol label switching (MPLS) traffic engineering capabilities, Ethernet yields quality controls only in terms of service classes, not the kind of per-session QoS controls offered by ATM constant bit-rate (CBR) guaranteed service or the constant information rate (CIR) guarantees offered via frame relay.

To achieve acceptable throughput for demanding real-time or bandwidth-hungry voice, video or transaction applications, the tendency can be to overbuild Ethernet network capacity. This strategy may provide enough head room to guarantee bandwidth, but at the prohibitive costs associated with often largely unused capacity.

As a more economic and efficient alternative, the Ethernet revolution requires tools to “right-size” the network to deliver resources where and when they’re needed. Specifically, Ethernet equipment should employ more than mechanisms such as DiffServ packet tagging to prioritize applications or users into class of service (CoS) queues. This approach ensures one train car will arrive sooner than another, but not that each passenger will arrive at the time promised.

In contrast, true per-session QoS mechanisms address not just bandwidth but delivery of each session as measured against that specific application’s latency, jitter, availability and other requirements. Such resource management also should be available on demand. If Ethernet QoS is to match ATM QoS, for example, it must be able to provide burstable, or variable bit-rate (VBR), as well as CBR transport, end to end--say, for an urgent storage backup or a one-time CEO video multicast application.

Ultimately the test of Ethernet service will be its ability to enable SLAs incorporating guarantees based on granular, per-session latency, jitter and packet loss metrics. This ability is the expectations benchmark already achieved by legacy ATM.

Even where Ethernet equipment supports interworking with legacy TDM, frame relay and ATM networks, it also must mature to replicate the fault-tolerance and non-stop redundancy capabilities of those legacy services. This capability means sub-50-millisecond re-routing around failures and in-service equipment upgrades--another set of benchmarks established by legacy communications networks. Better yet, customers should be able to demand in-service equipment software upgrades. Without these features, no QoS mechanisms can reverse downtime, resulting in poor service to customers.

Service providers should be aware that any Ethernet solution that does not incorporate deterministic QoS capabilities will hit a wall in satisfying demanding enterprise customers. Similarly, Ethernet WAN solutions without true QoS and non-stop uptime will not enable service providers to develop customized services.

The industry has only begun to standardize a way for Ethernet to seamlessly interwork with ATM and other transport protocols that offer per-session QoS for demanding enterprise applications.

Rare will be the enterprise customer willing to throw out functioning legacy frame relay, TDM or ATM gear to meet the provider’s timetable for revolutionary change. Customers must be able to continue leveraging the reliability and QoS of legacy solutions while introducing Ethernet on a phased basis, such as for new business locations.

In an era of heightened consolidation, a large bank or other enterprise may acquire a mix of Ethernet, frame relay, ATM and TDM access. This customer’s primary need, both immediate and long term, is for seamless, multi-service networking from any site to any other site regardless of access protocol.

The most flexible Ethernet solution will offer seamless Ethernet over fiber, Ethernet over Lambda, Ethernet over leased Line, Ethernet over DSL, Ethernet over IP/MPLS, Ethernet over Copper and Ethernet over Sonet/SDH.

Indeed, any Ethernet solution limited to fiber access will miss literally millions of enterprise branch offices and small and medium business locations. Deployed copper remains a valuable asset because so few businesses enjoy fiber connectivity and trenching requires high costs and long lead times. Consequently, success for Ethernet equipment requires not only multiservice interworking, but the ability to deliver services over fiber and copper. The latter is delivered typically through circuit-bonding techniques that transform multiple T-1s into one higher bandwidth connection.

The customer who is empowered to introduce Ethernet on its own terms is more likely to remain a customer.

So too will the customer who is asked to make no compromises on mission-critical QoS guarantees.

Michael O’Malley is a Marketing Manager in Tellabs’ global portfolio marketing group.

Visit Tellabs online.