By O. Ryan Tabibian
How much network bandwidth is enough? Here is a guide to the contending technologies in the deployment of high-speed networks--and their availability.
With the constant growth in desktop computing power and the rising popularity of bandwidth-hungry applications such as World Wide Web browsers and video conferencing, many IT suppliers and users long for the "wide pipes" that can carry large amounts of data faster than ever before. Such bandwidth, however, has been scarce and costly when available. A variety of solutions promise higher and higher bandwidth at lower and lower cost, but many of these solutions have been held up by physical and technical limitations. Finally, that is changing.
Existing technologies, such as Integrated Services Digital Network (ISDN), frame relay and leased lines (T-1 and the like), are growing in popularity and, more importantly, in availability. Other emerging technologies, such as Asynchronous Transfer Mode (ATM), Asymmetric Digital Subscriber Line (ADSL) and cable modems, are starting to appear.
Not all of these wide-area network (WAN) technologies compete directly with each other. ATM, frame relay and leased lines provide high-speed bandwidth for large corporate WANs. ISDN, on the other hand, is a more flexible technology that can provide both remote connectivity for corporate offices and Internet connectivity to the home. ADSL and cable modems are considered consumer technologies with possible applications in the commercial world. Similarly, these technologies often are at different stages of implementation readiness.
The path toward the wide pipe starts at the local-area network (LAN). While shared Ethernet to the desktop is still the network of choice for the majority of companies, many are looking to move to higher-speed technologies, with a combination of Fast Ethernet, Fiber Distributed Data Interface (FDDI) and even Switched Ethernet. In focusing on wide-area connectivity, however, the technologies noted above come into play.
For the past few years the busiest buzzword in the networking market has been ATM. Many observers have been pushing its merits as the ultimate solution for the future. However, while ATM has been the focus of the trade press, more readily viable technologies, such as ISDN and frame relay, have come to market. Other emerging solutions, such as ADSL and cable modems, while limited in availability today, hold promise for the future. (These solutions might not become commonly available until the end of the decade.)
The stakes are high for suppliers of these technologies. Some companies have bet their futures on the popularity of ATM, in both LAN and WAN markets. On the other hand, Regional Bell Operating Companies (RBOCs) point out that the only viable solution today is frame relay and ISDN. Cable and telephone companies tout cable modems and ADSL, respectively. Sorting out the proper choices requires some study.
ATM is a cell-based technology that uses continuous bursts of fixed-length 53-byte cells to transmit data, with a variety of network speed options that range from 25 megabits per second (mbps) to 622mbps. It is the only transmission technology both to allow simultaneous delivery of voice, data and video on a single access point and to guarantee delivery. This high-end technology is targeted toward organizations that require very high throughput or large network providers that need a flexible, high-speed solution for their backbones.
The main problem with ATM has not been its throughput but the lack of reliable, widely agreed-upon standards. A variety of standards are still being ratified in two important committees, the Europe-based CCITT, which develops communications standards, and the ATM Forum, a consortium defining and promoting ATM. Competing standards from vendors currently are being ratified in these committees. This means that there is no guarantee of interoperability among varieties of ATM equipment.
As in other areas of the IT industry, however, this bottleneck has not stopped vendors from pushing their solutions. While some vendors, such as IBM and Fore Systems of Warrendale, PA, are touting it as a LAN solution, ATM is more likely to find its niche as the backbone choice of network providers. MCI and others have embraced ATM by investing in retooling their infrastructures with it.
The MCI backbone, which was recently upgraded from an Optical Carrier (OC) 3 (155mbps) to an OC12 (622mbps), serves as the nerve center for InternetMCI and other MCI services. (OC is the optical characteristics of the signal that carries payloads.) While MCI offers a variety of options for connecting to the backbone, such as leased lines and frame relay, the backbone technology is pure ATM. "Frame relay did not offer the necessary bandwidth for our needs, and ATM has proved to be reliable," says Jack Waters, senior manager of Internet engineering in Washington, DC.
Another reason for MCI's move to ATM was flexibility. Unlike dedicated leased lines, such as T-1 and T-3, ATM allows for easier expansion of bandwidth. To avoid interoperability problems, MCI has deployed ATM switches only from Fore Systems.
Other providers also have found ATM to be a viable solution for their needs. Diamond.Net, an Internet service provider (ISP) and network integrator based in St. Louis, provides high-speed network connections to the Internet via its managed private ATM network, which consists of DS3/OC3 fiber optics and satellite connectivity. "With ATM we can also offer frame relay-integrated video, voice and data over a single type," says Andrew Gladney, Diamond.Net president and CEO. "ATM is the protocol that allows you to address all of the demanding and company-specific networking solutions, which are going to be demanded by a variety of institutions in the next few years."
ATM has also found a home in specialized applications, such as direct video feeds and multimedia development. Many research organizations, such as test labs, have deployed ATM to test and develop high-end multimedia applications.
As its name implies, this frame-based technology relays frames of information across a packet-switched network. The technology is more efficient than other technologies such as X.25, because it only performs error calculations at the source and destination, instead of at each switching node. This results in higher overall throughput. Frame relay is ideal for organizations that are looking for a less expensive solution than leased lines and don't need the high throughput offered by ATM.
Frame relay is generally affordable, thanks to the committed information rate (CIR) strategy developed as part of the frame relay spec. With CIR, you pay for only the bandwidth you use. It is like sharing a leased line among multiple users. For instance, while your connection might support 56kbps, you can have a CIR of 16kbps. And because the line supports higher throughput, in many instances you can burst above the CIR. Frame relay providers point out that this burst bandwidth is not guaranteed. If the backbone becomes congested, they automatically scale your throughput back to your CIR.
Some frame relay vendors scale back the traffic even if the network is not congested. When choosing a frame relay provider, make certain that it allows burst mode activity and will not scale your traffic back every time you exceed the CIR.
Another benefit of frame relay, aside from cost, is availability. Concert, a joint venture between British Telecom and MCI Communications based in Reston, VA, is one of the largest providers of frame relay. Currently, it provides frame relay in over 40 countries and plans to add more. Frame relay also offers much higher throughput than the reliable, albeit slow X.25, which mainly runs over 9,600bps lines. From a technical perspective, frame relay could be considered the logical next step for anyone who wants to upgrade from X.25.
DHL Systems, the package carrier, has chosen frame relay to interconnect its global network, DHLnet, which spans over 120 countries. Using this network DHL can provide customers around the world with package tracking and pricing information. Much of this network, which was previously based on X.25, was upgraded last year as part of a plan to decrease cost and increase bandwidth. DHL chose Concert as the frame relay provider, and now over 90 percent of its traffic runs over the frame relay network.
According to Doug Paustenbach, project manager for network services at DHL in Burlingame, CA, "ATM was not available internationally, and the price/performance was much better with frame relay. Frame relay is billed at a flat monthly rate, as opposed to X.25, which was volume-transmitted. We've lowered our global communication cost by 50 percent, while the traffic has doubled."
The shared bandwidth that makes frame relay affordable is also one of its problems. Having no delivery guarantees, frame relay is not designed for voice or video, which require guaranteed delivery. However, technologies on the horizon, such as Resource Reservation Protocol (RSVP), are intended to enable reliable voice and data transmission on frame relay circuits. RSVP allows you to reserve bandwidth between two Internet Protocol (IP)-based devices. A number of vendors, such as networking giant Cisco Systems of San Jose, CA, will be implementing this technology in future versions of their router software.
ISDN, which has been called the last great switch technology, has been around for some years. Until recently it suffered from limited availability in many parts of the U.S., and even today, while ISDN is common in major metropolitan areas, its availability is still limited in more rural sections of the country.
Unlike ATM and frame relay, which have no consumer applications, ISDN plays in both commercial and consumer markets. An established technology for remote site connectivity, it has also recently garnered popularity for Internet connectivity for consumers.
ISDN is a proven solution for point-to-point connections. Its bandwidth ranges from 56kbps to 128kbps for two-channel bonding. In the past year, the availability of equipment that supports bonding of two B channels in ISDN has also increased. With bonding you can combine the B channels to get an aggregate throughput of 128kbps.
ISDN also can serve as a backup solution to other technologies such as frame relay. DHL, for instance, is probing the possibility of using international ISDN as a backup solution. "We are looking at ISDN as a backup technology for the future or even as an enhancement to the current frame-relay network," says Paustenbach.
A related application for ISDN is switch access, which enables you to use an ISDN link to dial into a local frame-relay cloud rather than other ISDN lines. In this way you can make the ISDN connection in your home and become another node in the frame-relay cloud and thus an extension of the company's network and the Internet.
But ISDN is not without problems. Interoperability concerns remain among different ISDN switches. For example, ISDN switches between an RBOC and AT&T may not operate properly. The problem becomes amplified when deploying ISDN internationally.
Bandwidth is another limitation of ISDN. Competing technologies are poised to surpass ISDN's maximum throughput of 128kbps in the near future. However, while some, such as ADSL, promise higher bandwidth, their availability is still limited. ISDN will continue to hold a lead in availability for the next couple of years, before other emerging technologies begin to catch up.
Taking the place of ATM as the next big hype, cable modems and ADSL are the new promised technologies. The primary target for both is the home consumer market for high-speed Internet access, while ATM is aimed mainly at corporate networks. Today neither of these technologies is commercially available, except for a few pilot sites, nor will they be widely available before the end of the century. There is likely to be a variety of commercial applications for them in the future.
Cable modems, which run network traffic on existing cable wiring, hold the most obvious promise. Thanks to the push by cable companies, over 70 percent of all homes in the U.S. are wired for cable. Most of that cable is designed for analog television channels, which run at 8MHz. This means, under ideal circumstances, a single coaxial cable in the home can achieve 10mbps of throughput.
The most visible pilot work is being conducted by @Home, a joint venture between Tele-Communications, Inc. (TCI), of Denver, the nation's largest cable supplier; Comcast Corp., a telecommunications company based in Philadelphia; and Kleiner Perkins Caulfield & Byers, a venture capital firm based in Menlo Park, CA. The startup is conducting two pilots, one in Fremont, CA, and the second in West Hartford, CT. No results are available yet.
Unfortunately, there are a variety of problems with cable modems. First, most cable companies will have to retool to support the enormous push for digital data. That is, they will have to create the switches on the back end and get access with high-speed lines to the Internet. They will also have to install repeaters near each network segment.
The second problem is the chance for radio frequency interference with cable lines, which could cause packet loss and decrease throughput. Since cable companies, at the time of installing the cables to the home, did not envision the possibility of data, many of these installations are not up to spec.
Finally, unlike switched technologies such as ISDN and ADSL, cable modems are shared networks. For instance, the current design calls for sharing 30mbps of bandwidth per neighborhood, which can span to as many as 2,000 homes. Since there is little control over the usage of this bandwidth, a few aggressive users could occupy the majority of the network throughput.
ADSL is the telcos' response to the cable modem. It runs over standard telephone lines and promises throughputs of up to 8.448mbps download and 640kbps upload. But the actual throughput can vary depending on the distance of the line to a local loop. To achieve the promised throughput of 8.448mbps, the termination point has to be within 15,000 feet of a local loop, a limitation much more stringent than for ISDN.
Although phone companies are marketing ADSL as both a consumer and a business solution, some observers think that, due to the distance limitations, the technology will work only for businesses in metropolitan areas because of the higher density of switches. "ADSL is more competitive with T-1 or Primary Rate ISDN than as a consumer solution," says Ed Goldgehn, vice president and chief technology officer of Open Communication Networks, a network integrator and ISP based in Atlanta.
Obviously, each of the above technologies has benefits and drawbacks. Choosing the right one can be complicated. Purchasers should consider several factors before making a final decision.
First, evaluate your current network. It is critical to avert any bottlenecks that might exist internally before spending thousands of dollars on a new technology. Begin by creating some data models, and use a network analyzer to get a handle on your network utilization. For a more accurate assessment, use a network modeling application, which can provide a fuller model of your network.
Second, identify your bandwidth requirements. When it comes to determining bandwidth, the application should be the measure. Consider both current and future network applications. For example, if e-mail is the only application, an ISDN line should satisfy your bandwidth needs. However, if there is a need to pump voice, data and video through the network, with guaranteed delivery, ATM is the only solution. If you are just looking to speed up access to the Internet, consider leased lines.
This decision also depends on the network topology and the number of sites that must be interconnected. For instance, if your WAN topology is designed to connect multiple locations, consider frame relay, which allows you to connect multiple sites quickly and efficiently. However, for point-to-point connection leased lines might be the best solution. T-1 pricing is based on distance: the longer the distance between the two sites, the higher the cost. Again, evaluate other options, such as ISDN and frame relay, before investing in leased lines.
Third, it is important to define a realistic timetable for deployment. Do you need more bandwidth today, or can you wait for future technologies to emerge?
Finally, consider the cost. Getting access to ATM networks can be expensive, and the equipment can be even more costly. Solutions such as T-1 and T-3 carry large upfront installation and equipment costs, along with high monthly costs.
Overall, the variety of technologies currently available offers users a plethora of choices in wide-area connectivity. With the rapid increase in the availability of technologies in the future, consumers and companies alike should see more bandwidth for fewer bucks. The pipe may not be a dream after all.
O. Ryan Tabibian is a principal of Tabibian & Associates, a systems and network integrator based in San Francisco. He can be reached at ryant@best.com.