The Great Communications Race: Infrastructure Evolution

28 Mar 2003

There are two general media that carry information through networks: Electrons flowing in a metallic medium and electromagnetic waves propagated through space or through optical cable. Electrical signals and guided light travel in cable networks, unguided light and radio waves travel through space. All of the media can transport information in either analog or digital formats. 1

In each medium, digital information primarily is transmitted (“carried”) in bits: a string represented by zeros and ones. Each technology has a “carrier” that can be used in different ways to create two situations or “states”: one state representing the number 1, and another state representing the number 0. For example, in optical fiber, the carrier is the presence of light, often of a specific color, forced through glass by a laser or light-emitting diode. One of many ways to create the zero-one states is to assign two different pulse widths to the zero and one values. By mixing flashes of light having two different pulse widths, a receiver can sort out the zero and one bits in much the same way Morse code allows the construction of words from long and short flashes of unguided light, say from a flashlight or beacon.

The Public Switched Telephone Network (PSTN) 2 evolved from an analog network into a digital network. Originally the PSTN connected telephones by establishing a continuous circuit between the calling and the called parties and encoding a voice conversation in analog form. This technology dedicated the entire capacity of the circuit between telephones to a single call.

The efficiency of the PSTN has improved with the advent of digital technology. Packets of digital information originating from, and destined to, different locations can share the same infrastructure, much like cargo in boxcars on the same railroad route by assembling and disassembling along the way. 3 Nevertheless, in contrast to the Internet and related data networks, the PSTN still operates by establishing a virtual circuit (a specific path) between locations and maintaining that path for the duration of the call. An analogy is the tagging of bags at the airport: the specific flight numbers that the bags (hopefully) will ride from origin to destination are determined in advance and printed on the baggage tag.

TCP/IP-based data networks evolved in a different environment. These networks were conceived and engineered to connect computers in an environment where any particular path could malfunction. In the baggage tagging analogy, the bag would carry a destination tag but no particular flight numbers that must be used; the bag would be sent to intermediate airports that are capable of routing it to other airports that eventually connect with the destination. These networks use a routing protocol TCP (Transmission Control Protocol) and an addressing protocol IP (Internet Protocol) 4 to conduct communications. 5

Both the PSTN and TCP/IP-based networks can communicate using “packets” that ride the selected medium. A packet is a formatted framing of bits that can be interpreted by the equipment connecting segments of the medium. Just what bits are framed, how they are framed, and what transmitters and receivers do with the frames are what differs among networks. As mentioned earlier, the PSTN establishes a specific path for each call. TCP/IP-based networks allow packets associated with a single call to travel different paths from origin to destination, and to arrive in a different order from which they were sent. The packets of a call are assembled at the destination in the proper sequence based on special information that is encapsulated in the packet along with the information being communicated. The Internet is a classic example of a data network system based largely on TCP/IP. In order to understand the advantages of the Internet over the PSTN (and vice versa), a history of the two networks is useful.

The Internet, in principle, consists of a “web” of interconnected data networks that have been linked together in order to allow for the mutual exchange of data. The concept of a data network and the Internet evolved together and had common origins. In the early 1960s, packet communications and the idea of connecting two or more computers developed independently and in parallel within three scientific communities: MIT (working with ARPA – Advanced Research Projects Agency), RAND, and NPL (Btitain’s “National Physical Laboratory”). In 1965, the first, albeit small, wide area computer network was established when a computer in California was connected to a computer in Massachusetts. The connection was made using the circuit-switched telephone network. The inefficient connection served to strengthen the growing conviction that packet communications was the only viable way for computers to efficiently communicate. Thus, the Internet was born in principle, in concept and, on a very limited scale, in practice.

In 1968, the first comprehensive data network, ARPANET, was designed and funded by ARPA based on the packet-switching concept. Packet-switching was such a key element of the project that the initial step in building the network was to commission the construction of the first packet switch. Packet-switching became a central component of the network design because the competing network architectures required that each network be a subservient to a single “master” network. Only packet-switching would, in theory, allow one network to act as a peer of another. A fundamental principle behind ARPANET was that individual networks would vary in function and design for specialized purposes but might need to share computing capacity and data with other networks or computers. Thus, no master network would assume control of other networks, but each would act as a peer to the others when exchanging data. 6

A critical point in the development of the Internet occurred when the Packet Radio Network and the Packet Satellite Network were being designed to interconnect with ARPANET, a network then laced together with cables and wires. Radio communication was vulnerable to jamming, “dead zones” and other interruptions. In order for radio networks to connect to other networks without disabling end-to-end communications, the concept of “open-architecture” networks evolved. In an open-architecture model, networks need not conform to any particular architecture. The “internetworking” designers needed to ensure that lost packets would be retransmitted by the originating computer and that routing of packets would dynamically respond to temporary or permanent failures in any portion of the interconnected networks. The result was the development of the TCP/IP standard that is used today to address and transmit packet data through interconnected networks.

Thus, the Internet grew out of internetworking requirements that would tie together the first three significant networks: the ARPANET, the Packet Radio Network and the Packet Satellite Network. Today, the Internet allows virtually any arbitrary data network to communicate with any other provided that the two networks are connected and permit data to be exchanged..

The PSTN was designed to facilitate voice communications, a narrowband form of communications. Because two-way real-time communications was required, the PSTN was structured to keep a clear channel open between the calling and called party. As digital communications evolved, the PSTN was made more efficient by using the tiny periods of “silence” that take up most of the available capacity on a voice call interweaving other calls along the same communications path.

The most common means of “digitizing” voice calls today employs a technology called Asynchronous Transfer Mode (ATM). ATM switches package data into small fixed-size cells (packets) and sends them along a specific circuit that has been identified for the call. Many calls can share the same physical infrastructure because ATM switches can keep calls separate by uniquely labeling the packets that belong to each call.

The race between technologies for carrying basic telephony can be illustrated using ATM and Ethernet, a TCP/IP-based network. 7 A brief description of the advantages and disadvantages of each type of network will shed light on the likely winner. 8

During the past couple of years, bandwidth scalability of Ethernet and ATM have tended to rival one another. In the year 2000, gigabit Ethernet was well established with 10 gigabits on the horizon (today additional orders of magnitude increases are promising). ATM offered just under ten gigabits per second (OC192). Ethernet can easily be fragmented into one-megabit pieces (or even fractions thereof). ATM can be fractionalized as well but at a higher cost. In general, neither technology clearly dominates the other in scalability. Advantage: neither.

ATM networks are typically more reliable than are Ethernet networks since they are “hardened” to commercial carrier standards. ATM networks, for similar reasons, have better billing and accounting systems. The one advantage TCP/IP has over ATM in these regards is greater customer control to enable and control of custom features. The advantages of ATM are minimal in a private Intranet but can be significant in a public network. The limited control over features is a disadvantage of ATM. Advantage: ATM.

Looking to the future, one must consider that other communications infrastructures are becoming economically viable. For example, satellite alternatives to cable and fixed wireless networks are technologically viable (indeed, such a network was selected to network libraries throughout the state of Montana). Specialized wireless networks are being designed and engineered for various bandwidths over varying distances (witness the recent development of “hot spots” that offer access to the Internet in public places such as coffee shops). Future networks will need to be sufficiently flexible to adapt to these technologies as they evolve. Ethernets generally have been more flexible in adapting to alternative infrastructures than have ATM networks. Advantage: Ethernet.

Fixed wireless extensions of Ethernet and ATM networks use different equipment. For example, wireless bridges can extend Ethernet from nodes equivalent to central offices to customers’ premises. In a circuit-switched wireless network, extensions from central offices to customers’ premises can employ local telephone or cable facilities such as cable modems or DSL (Digital Subscriber Line) equipment. In general “extranet” extensions are less expensive and more rapidly deployed in an Ethernet network than in an ATM network. Advantage: Ethernet.

ATM can provide guaranteed throughput and less latency than can Ethernet. However, in an Intranet (a private data network), Ethernet can be equal the performance of ATM networks. Public IP networks however cannot yet match the performance of private networks. Advantage: ATM.

TCP/IP networks such as Ethernet are closing the gap in the race against the PSTN. As Voice over IP (VoIP) advances, the gap could be closed soon. For now, ATM and related forms of data communications 9 have the advantage in traditional telecommunications. However, telecommunications is increasingly being bundled with more advanced services such as Internet access, Personal Digital Assistants (PDAs), navigation systems, and other useful features. These bundled offerings favor TCP/IP networks. As communication devices migrate to a wider variety of infrastructures (satellite, household LANs, etc.), TCP/IP networks such as Ethernet that use open and widely understood standards encourage the development of compatible devices. Advantage: Ethernet.

We have concentrated on bit framing and protocols but there are other factors that affect the quality of voice communications. Two of the most important are: 1) the stability of the evolution of the technology, and 2) the frequency with which components of integrated systems are changed.

The Stability of the technology

Ethernet systems change much more frequently than do the carrier systems that drive the PSTN. Thus Ethernet equipment and engineering designs become outdated more quickly. This reduces the appeal of VoIP telephony over Ethernet (or other data networks) in companies and individuals who do not wish to change out or upgrade equipment and associated software frequently. Carrier systems in the PSTN remain more stable for longer periods of time. 10

Changing Operational Systems

VoIP and data networks often are integrated in the same infrastructure. Data networks like Ethernet are being constantly updated and modified in most business environments. These changes frequently affect the stability and reliability of the telephony application.11

In general, VoIP system failures tend to be more related to the frequency of human intervention than to the nature of the underlying technology. Advantage: ATM.

On balance, this author bets on TCP/IP-based networks closing the gap and taking the lead in the technology infrastructure component of the race within the next three years. Watch for progress in billing and accounting systems in TCP/IP networks. Watch for improvements in quality in VoIP. Watch for better protection from failures in VoIP due to human intervention in integrated data systems. And watch for more customer-controlled feature management in ATM networks. These developments could tip the balance.

Next week we discuss some of the economic forces shaping the race: economies of scale and scope, network costs, and the economics of open standards v. proprietary engineering.

1. Analog communications is accomplished by encoding information in the form of continuously varying values such as voltages or frequencies. Digital communications encodes information in discrete values, most commonly in strings of ones and zeros. Most communications networks today transport information in digital form for at least part of the journey. Since advanced communications networks today are digital, we will concentrate on that technology.

2. The PSTN generally refers to the traditional telephone network connecting telephones to a central office at which calls are switched through the vast network of cables and wires to reach other wired telephones.

3. The most common circuit-switched digital network technology used in the PSTN is Integrated Services Digital Network (ISDN).

4. IP deals with the formatting and addressing of packets while TCP deals with how computers establish communications and exchange information. Together, TCP/IP can set up and sustain communications between two or more properly equipped communications devices.

5. The most common such network is Ethernet, originally designed to connect devices located near one another (e.g., a computer to a printer or a mainframe to a mini-computer) but increasingly is used to reach across large areas spanning hundreds of miles.

6. A common misconception is that ARPANET (and therefore the Internet) was motivated by the cold war and the need to preserve communications in the event of a nuclear attack. While survivability was a deliberate feature of the ARPANET design, the design was not motivated by fears of a nuclear war. The first mention of ARPANET’s ability to survive such an event was in a publication by RAND that mentioned this advantage as ARPANET was evolving in theory.

7. These networks can exchange traffic easily today – the use of the two networks is not mutually exclusive. However, one “species” of network likely will dominate the other over time, thus the technology race one that likely will be won by one or the other (or their future offspring).

8. Before jumping to any conclusions, consider the impact of other forces to be discussed in later articles in this series. Technological superiority does not always prevail as the race between VHS and Beta videotape formats illustrated.

9. Frame Relay, for example, is another circuit-switched communications platform that differs from ATM in that it uses variable-sized packets rather than fixed-sized packets. Frame Relay is more suitable for voice than for video and data communications as a result.

10. As the PSTN is increasingly privatized and deregulated, depreciation rates are likely to shorten and competitive pressures may demand faster implementation of new features and services. Thus, this advantage of the PSTN over TCP/IP networks may decrease.

11. Examples of activities that often disrupt VoIP on private networks are: changing encryption Internetworking operating systems; IP subnetting updates; changes to VLANs,; and so forth.

Dr. Richard Emmerson, Founding Director – PriMetrica