The fiber evolution: from backbone to building

Th 0104lwspr04f1

Premises Networks

Technologically proven in the backbone, fiber is now poised to become the media of choice for enterprise and residential customers.

DICK WILLSON, Allied Telesyn

Because of its superior integrity, security, and bandwidth capabilities, all indications are that fiber-optic media will be the media of choice for delivering communications services to both enterprise and residential customers.

Fiber-optic transmission technology was initially developed for long-haul, undersea communications between continents. Once the submarine cables were operational, intercontinental bandwidth increased significantly and tariffs started to fall as traffic migrated from satellite circuits to fiber-based undersea cables.

The submarine fiber-optic cable system initially required optical-electrical-optical (OEO) conversions for signal amplification, but with the development of erbium-doped fiber-amplifier (EDFA) technology, the OEO conversion requirement was eventually eliminated. Today, signals can be transmitted across the entire length of an intercontinental cable system without a change in form, meaning that the cables can move more data faster, thanks to the elimination of OEO conversion.

Contemporary terrestrial trans continental long-haul fiber-optic cable systems also use the EDFAs, making city-to-city links faster and more efficient, as well. This efficiency was enhanced even further with the development of DWDM, which allows information to be coded for transport on different wavelengths (colors) of light.

With all of the recent advances, there is still a lot of "dark fiber" out there today: long-haul intercontinental and terrestrial transcontinental fiber-optic systems are capable of delivering more bandwidth than we need. Bandwidth is derived from the fibers that are lit at any given time. In fact, if all of the fiber strands in today's fiber-optic cable were lit, there would be a glut of bandwidth; but as it stands now, there remains untapped capacity.

The telecommunications industry is not the only user of fiber-optic technology. Large enterprise campus sites use fiber technology to implement LANs. Compared to the telephone companies' long-haul networks, LANs are very short, so there is no requirement for optical amplifiers. Fiber-optic backbones are frequently used for the interconnection of high-speed LANs using Ethernet and Internet protocols. These high-speed backbones typically move information at 1 Gbit/sec today-a speed that will very soon increase to 10 Gbits/sec.

Today, copper twisted-pair is the predominant medium used by the incumbent telephone company for the delivery of communications services to enterprise and residential customers. Plain old telephone systems have been the primary means of communicating both locally and long distance. The problem is that the local loop was designed for the transmission of voice communications. It's a mature technology, but inadequate by design-the amount of bandwidth that can be delivered is restricted by the characteristics of the copper twisted-pairs installed between the customer and central office. Fiber in the local loop-even to business premises-reportedly accounts for less than 1% of current connections.

There is a lot of interest in the deployment of fiber-based metropolitan-area networks (MANs), which provide the connection between customers in a given city to customers located in another remote city via the long-haul fiber-optic transcontinental network. MANs also provide the connectivity among customers within a city. An effective, efficient MAN requires the deployment of fiber-optic technology within city environments.

Because fiber-optic technology has the potential to deliver an enormous amount of bandwidth, some commentators have promoted the notion that optical-networking advances will eventually make bandwidth "free." In reality, we all know that nothing is truly free. In the United States, residential telephone customers have been educated to believe that their local telephone calls are free, when in reality, pricing is simply set so that for a fixed monthly fee-the monthly basic service charge-telephone customers can make an unlimited number of local voice calls. The same tariffing methodology can be translated to optical networking whereby, for a fixed monthly fee, we are always connected to the Internet. To translate that to mean "free bandwidth," is as misguided as the notion that local calls are free.

Today, multiple technologies are used to deliver communications services to customers. There is the legacy telephone company network services, fiber-optic technology in the long-haul backbone, Ethernet in the access loop, and fiber in the MAN. Fiber is an important component in this mix of technologies. To deliver services to customers, an understanding of optical technology and how the various technologies interwork with each other is critical. Th 0104lwspr04f1

Figure 1. The bandwidth speeds supported by LANs versus those within the incumbent local-exchange carrier infrastructures are totally mismatched, resulting in today's access bottleneck.

The dominant, somewhat monopolistic, local service provider is the incumbent local-exchange carrier (ILEC). It is impossible to ignore the immense infrastructure and financial strength of the ILEC, especially in comparison to the small fledgling competitive local-exchange carriers (CLECs). The ILECs own the local copper twisted-pair cabling that was built over the years, primarily to support voice telephony. Those same ILECs have been installing an increasing amount of fiber-optic cabling in the local loop to support the local voice telephony infrastructure and provide interconnection with the long-haul fiber backbone.

The ILEC service available to customers follows a strict hierarchy determined by the voice telephony infrastructure. Customers' LANs today support 10/100/1,000-Mbit/sec bandwidth, but the ILEC's voice infrastructure provides 1.544/45/155/622/2,500 Mbits/sec-a total mismatch (see Figure 1). The delivery of services via fiber to customer premises, where customers can choose the bandwidth they require on demand, is still in its infancy.

Optical-fiber links use DWDM to create multiple wavelengths, or light paths, of individual colors. In practice, each light path is a dedicated point-to-point circuit of a given bandwidth. DWDM technology is generally used in the long-haul backbone.

In the access network, each Ethernet packet carries addressing information. The network nodes examine each packet to make the routing decision for the next hop. In a DWDM network, the wavelengths are set up prior to any traffic being accepted. Because the wavelength is determined before any traffic is accepted, a DWDM network never has to examine or process any packet; the information (content and addressing) is coded into the light pulses digitally and remains as light right through the DWDM network.

In networking jargon, Ethernet and Internet Protocol (IP) networks are connectionless systems, while DWDM networks are connection-oriented. In practical terms, if the DWDM systems had to examine every packet, the light pulses would have to be converted into electrical pulses to process the address information. The bandwidth of the DWDM system would then be limited to the processing speed of the silicon-based devices, severely limiting the bandwidth of the system.

Ethernet technology over the years has reinvented itself several times and is now the de facto LAN technology used in the enterprise for data communication. Ethernet will operate at 10, 100, and 1,000 Mbits/sec, and we're on the verge of a jump to 10,000 Mbits/sec this year. Ethernet has also "escaped into the local loop" and is proving to be the most cost-effective access technology, delivering multiple services (data, voice, and video) to customers. Its support for multiple transmission media-copper twisted-pairs, telephone wire, fiber, wireless and wireless optics-makes Ethernet a good choice for broad-based deployment.

A packet-based networking technology, Ethernet wraps information in packets together with source and destination addresses for passage to and from the user, and the IP suite defines how the packet is routed from its source to its destination. IP is the foundation of the Internet and consists of a robust set of protocols that enable the network to detect whether a node or link is unavailable, finding an alternative route to the destination automatically when necessary. If packets are lost, an IP protocol retransmits the original packets. IP also statistically multiplexes packets from different applications over the same link, using the bandwidth available on the communication link more efficiently.

Probably the most significant recent development is the establishment of a 10-Gbit/sec Ethernet standard (IEEE 802.3ae) this year. This is first time a LAN standard will support fiber-optic connectivity to both the WAN and LAN. It provides the standards for the interconnectivity between LAN and WAN SONET/SDH in the United States and Europe, dark fiber using multimode and singlemode fiber-optic technologies for DWDM transceivers as well as packets and circuits.

The largest global network is the voice telephony system. Even the Internet relies primarily on the telephone system for access (see Figure 2). The telephony system uses circuit-switch technology to interconnect users. The Internet is a data communication system that uses packet-switch technology to interconnect computers. As previously discussed, the DWDM fiber-optic technology used in the long-haul backbone creates wavelengths or light paths for point-to-point connectivity. Both Ethernet and IP use packets in the access loop.Th 0104lwspr04f2

Figure 2. The largest global network is the voice telephony system, providing access for today's Internet.

The consensus is that information (video, voice, and data) will be digitally coded and all communication network services will be digital. Ethernet and IP are the de facto packet-switching technology used by customers, the LAN, and the Internet. Since the Internet is a global network, when an end-to-end connection is made it must traverse two different technologies: a packet-switched network and an optical DWDM long-haul network. How much of this path is over a packet network compared to a DWDM network depends on the economics of each technology.

If it becomes economically feasible to install fiber-to-the-home systems, service can be delivered via several light paths for a circuit-switched network. A light path would connect the customer to a service, for example, streaming videos. The actual information would be digitally encoded and could be encapsulated in a series of packets. In this example, we are transporting packets over a circuit, much like today when using the telephone network to access the Internet, but at much higher speeds.

With the definition of the 10-Gigabit Ethernet (10GbE) standard, the networking industry has defined a very important interface between a packet-switched network, the Ethernet LAN or MAN, and the circuit-switched fiber-optic DWDM WAN backbone (see Figure 3)-again, making it possible to deliver services encoded as packets to all customers using fiber-optic technology. In this scenario, however, we have a packet-switched Ethernet/IP access network operating over fiber, with the interface between the packet-based access network and the circuit-based DWDM long-haul defined by the 10GbE standard.Th 0104lwspr04f3

Figure 3. The standard for 10-Gigabit Ethernet has defined a very important interface between a packet-switched network, the Ethernet LAN or MAN, and the circuit-switched fiber-optic DWDM WAN backbone.

Therefore, packets and circuits should always be seen as complementary-where the interface between each technology occurring in an end-to-end connection will be defined by the economic cost and business strategy.

As fiber-optic technology is deployed to build networks, both packet- and circuit-switched technologies will be used. Depending on the economics, the boundary between the two technologies will move closer or further away from the customer and be located in different parts of the end-to-end connection path. The specific business models to which service providers subscribe will ultimately determine the location of the interface.

Today, the largest global network is the circuit-switch-based telephone system, and the majority of services are still delivered by large national monopolies via this global network. We are seeing the beginnings of deregulation and privatization of the monopoly telephone companies, and regulators are striving to "open" the local loop for access by the competition. Because the incumbents are large organizations and opening the local loop is not necessarily in their own best interest, change is slow.

Customers connect to services via the local loop. Deregulation and privatization have attracted a number of competitors to vie for customers' business. The long-distance interexchange carrier has the advantage of extensive installed fiber-optic facilities, although not in the local loop. Cable-TV companies supply data services via cable modems chiefly through coaxial-cable infrastructure, which is being upgraded to fiber. The ILEC owns the copper twisted-pair cabling that dominates and therefore still has an effective monopoly in the copper local loop. The ILEC also has extensive fiber facilities that are not currently accessible to customers.

A full-facility CLEC would implement its own infrastructure using fiber as well as using copper twisted-pairs from the ILEC. Building local-exchange carriers (BLECs) implement a networking infrastructure only within a building and rely on other service providers for service to the building. There are others, such as the onsite service provider, similar to the BLEC. The Internet service providers (ISPs) and application service providers generally do not own any networking infrastructure. Finally, an Internet hotel provides collocation facilities.

Deployment of fiber-optic technologies in the MAN and local loop is extremely capital intensive. Reports estimate the cost of installing fiber in the MAN at between $300,000 and $500,000 per mile-70% of that cost is attributed to labor. The deployment of a network that can be a reasonable alternative to the local telephone company is going to depend largely on funding for potential competitors. Funding for the deployment of fiber will be derived from a variety of sources, including new competitors funded by venture capital as compared to the incumbents that typically fund capital expenditures from ongoing revenue.

So although the opportunities are huge, the incumbent carriers present formidable competition. Though regulators have assumed that opening the local loop to competitors would attract sufficient private venture capital to enable the deployment of alternative services to customers, only time will tell. Over the last few years' bull market, funding was out there-but now, with the bull in retreat, the realization that high-tech and Internet companies were distinctly overvalued is likely to affect the availability of funding. Telephone companies have been able to acquire competitors at extremely attractive valuation, creating an even stronger incumbent monopoly.

There are alternatives for the delivery of high-speed broadband services that are far less capital-intensive than building a competitive infrastructure that simply mirrors the incumbent phone companies' local loop. In some local communities, the need is strong enough and the economic justification sound enough to lead the community to develop "broadband islands" independent of the telephone companies. The technology to develop these islands already exists-the LAN.

Local community islands can be connected to each other and the Internet just as enterprise LANs are interconnected. What is required today is to encourage local communities to build their own LANs-essentially their own broadband islands. That's not a new idea; it is essentially how the Internet itself was first established. The telephone companies did not develop the Internet as a service for phone customers; the advanced research and academic community developed it by connecting building and campus LANs to communicate with one another and share information. These organizations developed IP to interconnect their local computer networks. They leased the necessary wide-area communications from the telephone companies to implement what was their private network.

In contrast, broadband services are built from broadband islands-LANs that are connected to other LANs. Initially, these broadband islands would use local phone services to access local ISPs' islands to use the Internet as the ultimate "inter-island" connection vehicle. Local telephone charges shared among the members of the broadband island community make the cost affordable to each member in the same way that the Internet was affordable to its pioneers.

Standard Gigabit Ethernet technology will form the infrastructure for the community island-it's already happening. There are communities where fiber-optic technology is being deployed by the city as a community resource. In these cities, dark fiber can be leased and the broadband islands interconnected using that dark fiber at gigabit speeds.

To deploy the broadband services described here is not capital-intensive because the members share the cost of implementing the local broadband island as well as the cost of the local telephone company access and dark fiber (if it is available). Communities in Canada, Scandinavia, and Brazil have already applied this concept to implement their shared broadband islands. In fact, a Gigabit Ethernet network is operational across the continent of Canada, interconnecting many broadband islands located in cities across the country.

Fiber-optic technology is the choice for networks in the future. The capital investment required to deploy fiber into the local loop is large, so organizations that want to compete for customers with alternative services will need substantial funds to be able to compete with incumbents.

There are alternative scenarios that are not so capital-intensive. Communities can deploy broadband services using the Gigabit Ethernet broadband island model. It is likely that the up and coming 10GbE will accelerate participation from companies in the LAN sector. When they begin to compete for business in the local loop, that competition should result in lower cost and lower prices, thus contributing to the possibility for widespread deployment.

Dick Willson is the chief technology officer at Allied Telesyn International (Bothell, WA).

More in Network Design