The need for capacity in metropolitan areas will continue to grow as businesses deploy more information technology and make further use of the Internet for e-commerce. As a result, established carriers need to be able to expand their metropolitan area networks (MANs), and new entrants need inexpensive technical solutions from which to launch their operations.

The appeal of Gigabit Ethernet stems largely from its simplicity and low cost. Extending its use into the local loop is appealing to carriers since it is easy to interface customer LANs that use Ethernet technologies with the network. The elimination of protocol conversions at customer premises makes Gigabit Ethernet cheaper than conventional solutions.

Most of today's network engineers have grown up with Ethernet. The technology is well understood and thoroughly standardized through the IEEE 802.3 committee.

However, the range of physical layer options included within the standard has multiplied from the original single coaxial cable-based system to an array of electrical and optical interfaces. It has increased in speed from 10Mbit/s to 1Gbit/s over the last 10 years. Work is now underway in the IEEE 802.3ae committee to deliver a 10Gbit/s option, and extension to 40Gbit/s is highly likely.

Although the general principles of operation of Ethernet have been maintained throughout its life, the physical network structure, speed and types of cable used have all changed considerably. Gigabit Ethernet can be supported over both copper and fiber. However, only short distances (less than 100 meters) can be covered with copper; this makes fiber, with a range of five kilometers, the only realistic choice for metropolitan area applications.

The core network of Gigabit switches is deployed around a city to provide coverage of the buildings required. They are interconnected using either Gigabit Ethernet links or some other form of transmission technology, such as dense wavelength division multiplexing (DWDM). Customers are connected to the nearest switching point through fiber. Local phone companies such as PacBell place their switching nodes in the basements of the office buildings in which their customers are located, rather than at an independent office location.

Not for everyone
Although established carriers can use Gigabit Ethernet in their networks, it is not yet the ideal solution for them. Most telcos and the older metropolitan carriers have extensive deployments of Sonet/SDH. This is not ideally suited to carrying packet traffic (since it provides rigidly channelized bandwidth), but it does offer two significant advantages from an operational perspective. First, the network can recover quickly from any problems caused by failures or fiber cuts. Second, network management standards and tools are available that allow the carrier to monitor and control the network to a reasonable level of granularity.

Carriers that adopt Gigabit Ethernet are making a leap of faith in adopting proprietary solutions that do not allow the degree of network control offered by Sonet/SDH. The risks are not high for a small network, but they become significant in the context of a major carrier deployment. Established carriers need the Gigabit Ethernet vendors to develop products that are more operationally compatible with existing technology.

At present, these weaknesses and the use of proprietary technology are preventing many established carriers from adopting Gigabit Ethernet. Carriers need solutions that integrate well with their existing systems at both transport and operational levels. The relevant standards should be completed by mid-2002. If these standards are successfully implemented, Ethernet technology could soon be ubiquitous.

Iain Stevenson is a principal analyst for next generation networks and infrastructure at Ovum, a consulting and market research firm.