Optical Ethernet service providers leave copper in the dust

Th 0101lwspr04f1

A new class of carriers offers Gigabit Ethernet over fiber, allowing LAN customers to bypass the local-loop infrastructure.

Yipes Communications Inc.

Over the last several years, dial-up networking has yielded to broadband technologies such as DSL and cable modems. These new approaches to providing bandwidth over existing copper lines establish a superior data link over the first mile to a business or residence at a reasonable cost. The name "broadband" is used to differentiate these last-mile technologies from the limited bandwidth attainable from dial-up networking. Both DSL and cable modem services, which have given residential and business users tremendous amounts of affordable, always-on bandwidth, are expected to grow exponentially over the next several years to reach millions of residential and business customers.

In as much as DSL and cable are a quantum leap over dial-up, these technologies are still inherently limited by the physical properties of the copper medium. Regardless of whether the delivery line is twisted copper wire in DSL or coaxial cable in cable-television systems, these media have physical properties such as ohmic resistance, frequency-dependent attenuation, crosstalk, impedance mismatches, and other limitations that place a relatively low bound on their ability to carry digital data. Furthermore, all of these media were originally installed for a different purpose than the transmission of data, and their adaptation to data in essence requires some backward engineering for data protocols.Th 0101lwspr04f1

Figure 1. Transporting data from an Ethernet LAN in one location to an Ethernet LAN in another location over a traditional carrier, using a voice-centric infrastructure, requires multiple protocol changes as well as substantial investment in different types of hardware to support the varied links.

Even services traditionally identified with data such as T1 (1.54 Mbits/sec), DS-3 (44.736 Mbits/sec), and frame relay are subject to the physical limitations of copper as well as the excess baggage of protocol conversion to carry data over historically voice-centric networks. Figure 1 shows the steps a traditional carrier using a voice-centric infrastructure must employ to provide a relatively simple connection between two Ethernet LANs.

Given an opportunity to build a carrier-class data network without any consideration of existing voice and local-loop infrastructure, the network would undoubtedly look much different.

In an optimal data solution, the network would use optical transmission from end to end, to remove the low-bandwidth bottlenecks imposed by copper. That means the end user must have access to the optical portion of the network, which is only possible if the optical network comes all the way to the customer's premises.

Secondly, the network must be scalable. One of the biggest obstacles today is the inability of conventional voice technologies to smoothly scale up several orders of magnitude to meet the wide range of bandwidth demands encountered in today's data-centric world. The network must be flexible enough to grow with the bandwidth needs of its customers and to do it on demand to meet fluctuating requirements.

The ideal network architecture is as simple and flat as possible. Converting protocols is complex, and the hardware needed to handle the conversion is typically expensive. Since the predominant protocol in use today at the customer's site is Ethernet, the obvious choice would be to use Ethernet from end to end.

The network must also be robust, meaning that at a minimum there can be no single points of failure. Business operations are tied to the availability of the networks that a company uses. Downtime means a loss of productivity and revenue. The all-too-common scenario of a construction crew cutting an underground cable and creating a prolonged service interruption can no longer be tolerated in an information-driven economy.

Finally, the optimal networking solution has intelligence and active management. Problems that arise must be analyzed and addressed promptly. The network must be responsive to the users, and it must have systems and a management organization behind it that provide customers with responsive information and support.

The network must do more than just provide data transport. It must support services that are of value to the customers that use the network. These services include encryption for security, firewalls, the ability to merge voice into data, and ultimately the ability to respond to software applications, which have different networking requirements. Th 0101lwspr04f2

Figure 2. Connecting two LANs with an optical network based upon Gigabit Ethernet technology that does not require protocol conversion. This network provides the foundation for building additional intelligent services such as encryption, firewall, or even application-dependent networking.

The technology to provide these services is available now. Companies have been founded to execute on these principles and deliver them as a service to the end user. It is important to note that to execute this well-the most cost-efficient (and hence most competitive) approach-requires that this architecture be built optimally with no legacy carryovers. Figure 2 shows the simplicity that can be achieved by building such a network.

There is no secret business case for providing such a carrier-class network service. There is, however, a very significant cost penalty if an existing network infrastructure tied to older technologies must first be cannibalized to achieve these goals.

How are the objectives of this ideal network realized in a world filled with copper? It starts at the metropolitan level where for several years dark fiber has been installed throughout business districts and concentrated commercial areas. The network service provider acquires access to this in-ground dark fiber from other carriers and lights it using Gigabit Ethernet technology. This fiber typically runs in a ring topology, which provides a natural basis for redundancy at the optical level.

Customers are connected to the fiber ring through the construction of short laterals. These laterals enter a building and "light" the building by bringing the optical carrier all the way to the telecom room, completely bypassing all the legacy copper infrastructure of the competitive and incumbent local-exchange carriers. The optical carrier is terminated in a wire-speed ASIC-based Ethernet switch, which can be provisioned as either a Layer 2 Ethernet metropolitan-area-network service or Layer 3 IP Internet access service. Regardless of the service, customers connect their LANs using Ethernet; there is no need for protocol conversion or telecom-based interfaces such as T1s, DS-3s, or OC-x. The ring topology ensures that even in the event of a fiber cut, traffic can traverse the fiber in the opposite direction. Although that sounds much like SONET, it's not-it's Ethernet.

Within a metro region, the network service provider aggregates many fiber rings in points of presence (PoPs). The PoPs consist of additional Layer 2/Layer 3 switches and routers, all of which are wire-speed, nonblocking devices with Ethernet interfaces. The aggregation points enable customers to logically connect Layer 2 on different rings within a geographical area and allow Internet-bound traffic to be routed through the most direct Internet backbone carrier to the ultimate destination. The design of the PoPs is fully redundant, as are the links to the Internet and links to the metropolitan rings. This redundancy, which is both physical and logical, ensures that the failure of any device, regardless of the reason, will not interrupt the ability of the network to carry the customer's traffic.

Unlike legacy access circuits, scalability of bandwidth over many orders of magnitude can be achieved using Ethernet framing. Rate-limiting capabilities of Ethernet hardware allow smooth scaling from Mbits/sec up to Gbits/sec, all of which can be done over a single interface with no change in protocols. On the near horizon are 10-Gbit interfaces. This increase in bandwidth will grow exponentially with the addition of optical DWDM technology, which will provide the ultimate solution in bandwidth scaling.

When a customer desires service at any given bandwidth, the network's provisioning system can deliver it almost immediately. Since the bandwidth and services a customer receives are dependent only on the logical configuration of the network, changes can be made with minimal delay and effort. There is no need for changes in the physical topology or hardware.

Provisioning is a complex task involving traffic engineering, configuration management, and links to back-office systems for billing and service management. But in a flat-topology optical network, which can be provisioned completely through logical changes, the entire process can be more easily automated. Provisioning and traffic engineering can also ensure that no fiber is ever oversubscribed. Thus, customers always have guaranteed access to their full bandwidth regardless of the traffic level on the network.

And finally, what about the advanced services such a network makes possible? This optical-network foundation supports the implementation of advanced service features such as security, encryption, firewalls, communities-of-interest, and the ability to dynamically adapt services to software applications.

Tom Kruckewitt is director of customer network integration architecture and Scott Reichmeider is senior technical consultant at Yipes Communications Inc. (San Francisco).

More in FTTN/C