acm-header
Sign In

Communications of the ACM

Home/Magazine Archive/June 1999 (Vol. 42, No. 6)/Technical Opinion: Does Data Traffic Exceed Voice.../Full Text
Communications of the ACM

Technical Opinion: Does Data Traffic Exceed Voice Traffic?

By A. Michael Noll
Communications of the ACM, June 1999, Vol. 42 No. 6, Pages 121-124
10.1145/303849.303871
Comments
View as: Print Mobile App ACM Digital Library Full Text (PDF) Share:

Mergers and acquisitions of enormous financial proportions have been redefining the media and telecommunication industries since the Internet began to emerge as a form of mass communication and commerce. The view that the Internet represents the future on a grand scale appears to motivate many of these megadeals. More and more good money from established telecommunication corporations is chasing Internet dreams, some of them based on claims that Internet traffic exceeds telephone traffic.

But claims of data and Internet superiority over voice in terms of network traffic are exaggerated and seem to be based mostly in folklore. I offer some simple estimates to argue that voice traffic (measured in bits) far exceeds data traffic and most likely will continue to do so for the foreseeable future (see the accompanying sidebar). The gap between data and voice traffic should not be surprising. A single-spaced typed sheet of paper contains about 30,000b of ASCII, which is equivalent in bits to only about 0.5 seconds of a telephone speech channel. Thus, a five-minute telephone conversation generates as much traffic—in bits—as 640 typed sheets of paper. Speech, though natural, is an inefficient means of communication in terms of bits.

The greatly exaggerated claims by some analysts that data will exceed voice are wishful thinking, stimulated by the media and stock market frenzy surrounding the Internet. When billions of dollars chase a limited number of high-technology investment opportunities, the investment climate stimulates exaggeration and hyperbole—and increasingly risky investment strategies.

The notion that data traffic exceeds voice traffic goes back to 1961 when Frederick R. Kappel, then president of AT&T, forecast that "...within 15 years or so, the volume of information communicated between machines may be even greater than the amount of communication between people" [2]. To refute this claim, R. Hough, while at Stanford Research Institute in 1970, showed through careful analysis that telephone traffic would exceed data traffic for the foreseeable future [2]. I confirmed the conclusion that telephone traffic would continue to exceed data for the foreseeable future again two decades later [3].

Claims continue to be made by telecom analysts and executives betting billions on Internet business strategies that Internet data traffic either already exceeds voice traffic or will do so very soon—even this year. Some telecom analysts go so far as to predict the decay—and even the death of—plain old telephone service [1, 6]. Other telecom analysts claim data traffic will predominate under any Internet growth scenario [7]. But, as far as I am concerned, no real data or analysis has yet been presented to support these claims.

Back to Top

The Video Factor

Is there any way Internet traffic could exceed voice telephone traffic. The only way this could occur would be for video to be carried over the Internet. Video involves long holding times, typically measured in hours of television viewing. Video requires a bit stream of about 4Mb/sec when compressed. But from an engineering perspective, a packet-switched network, like the Internet, is no way to transmit television to consumers, due to the disjointed nature and address overhead inherent in packet switching. A broadcast network architecture using either radio waves or coaxial cable is much more efficient and far less costly. The failure by telephone companies to provide switched video in the supposed convergence of television and telephony shows how visions of the future go sour. I doubt consumers will be watching television over the Internet anytime soon.

The historically "free" Internet makes it difficult to determine true costs and corresponding prices to consumers. I estimate that a reasonable price for the Internet-backbone, packet-switched network would be 1 micro-¢/b (10−6¢/b) [5]. At this price, downloading a 2MB video file would cost 16¢. This cost is not prohibitive for a typical consumer downloading a movie one time, but done routinely every day would be costly. Video compressed at 4Mb/sec would be extremely costly at 4¢/sec, or $2.40/min, although discounts would certainly be available for transmitting such large volumes of data. From a financial perspective, it would be far too costly to use the Internet to download large files, particularly video. Buying a compact disc or even renting a tape from a store and watching it through a VCR would be much less costly.

Local access for Internet traffic is the real issue for phone companies. The local telephone switch is designed to handle voice traffic with relatively short holding times and continuous transmission. Data, however, has long holding times and bursty transmission. Converted into bits, local-loop data traffic would far exceed voice telephone traffic, due to the long holding times for data. But because much of the traffic being switched is zeros, the challenge for phone companies is how to separate data from voice traffic as early as possible at the local level. One high-bandwidth solution is an asymmetric digital subscriber line (ADSL).

From a technology perspective, the requirements for voice, data, and television are quite different [4]. The need for multiple transmission modes, rates, media, and financial incentives implies different network architectures for their delivery—contrary to the claims that telephone calls will soon all be carried over the Internet.

Back to Top

References

1. Gilder, G. The death of telephony. Gilder Tech. Rep. 2, 9 (Sept. 1997), 2.

2. Hough, R. Future data traffic volume. Comput. 3, 5 (Sept./Oct. 1970), 6–12.

3. Noll, M. Voice vs. data: An estimate of future broadband traffic. IEEE Commun. 29, 6 (June 1991).

4. Noll, M. CyberNetwork technology: Issues and uncertainties. Commun. ACM 39, 12 (Dec. 1996), 27–31.

5. Noll, M. Internet pricing vs. reality. Commun. ACM 40, 8 (Aug. 1997), 118–121.

6. Rockstrom, A., and Zdebel, B. A network strategy for survival. IEEE Commun. 36, 1 (Jan. 1998), 36–40.

7. Wohlstetter, J. Packet policies: Petabits, photons, phonemes, and the feds. New Telecom Quart. 6, 1 (First Quarter 1998), 3–13.

Back to Top

Author

A. Michael Noll is a professor at the Annenberg School for Communication at the University of Southern California in Los Angeles. His home page is at www.citi.columbia.edu/amnoll/.

Back to Top

Sidebar: Back-of-the-Envelope Estimates

Telephone speech (when digitized) creates 64,000b/sec in each direction of the call, or a total two-way digital traffic of 128,000b/sec. An email of 200 words requires about 9,600b. Thus, 11 such emails would be equivalent in traffic to only one second of two-way telephone speech. A one-minute telephone call generates the same traffic in bits as nearly 700 emails. If data traffic exceeds voice, it is clearly not because of email traffic.

What about Web browsing? Assume an average Web page represents 40,000B and that 20 such pages are browsed by a typical user in the U.S. on an average day. The data traffic generated through such browsing represents 6,400,000b, or the equivalent of the two-way traffic generated by a single telephone call of only 50 seconds. On an average day, I participate in well over an hour of long-distance calls. So I would have to browse 1,440 Web pages to equal the two-way traffic of one hour of long-distance calls.

These back-of-the-envelope estimates do not alone substantiate the claims by telecom analysts that Internet traffic vastly exceeds voice traffic over backbone data and voice networks. So, to gather more evidence to quantify data and voice traffic, I conducted an informal study in 1998 of the communication activity of my graduate and undergraduate communication students at the University of Southern California in Los Angeles.

To estimate traffic, I asked them how many telephone calls they made and Internet interactions they participated in on a typical day. They told me they made and received an average of 20 telephone calls and 20 emails; they also surfed the Internet and explored 20 Web sites per day on average. I then converted these forms of telecommunication traffic into a total number of bits per day (see Table 1) and detailed these estimates (see Table 2).

Telephone traffic represents considerably more bits than any other form of data traffic today. The closest seconds in overall number of bits are Web surfing and the downloading of large files, but even these activities represent only 2% of the traffic generated by telephone calls. Adding together all forms of data communication on a typical day in the U.S. gives a total of 1623105b, or roughly 2% of the telephone traffic.

The total number of emails and Web-surfing screens per typical day in the U.S. is 40, or twice the number of telephone calls on a typical day. Thus, the number of data interactions—measured as the total of all emails, Web sites, and files downloaded—exceeds the number of telephone calls on a typical day. Therefore, on a typical day, the number of Internet interactions exceeds the number of telephone calls. But this statement is a far cry from the claim that Internet data traffic (in bits) exceeds voice telephone traffic.

Caveats

These estimates of data traffic are probably somewhat high, since they are based on student use at a communication school emphasizing Internet literacy, and hence the estimates do not represent the overall U.S. population. Moreover, most of these students had full- or part-time jobs, and were thus heavy users of the telephone. But they also were technology literate and heavy users of email and the Internet. My guess is that my estimates of the data traffic would have to be reduced by a factor of four to represent the overall U.S. population, even though business people might be even heavier users of the telephone and email.

But back-of-the-envelope analyses are good tests of the sanity of more thorough analyses. In any case, my informal survey still supports my contention that voice telephone traffic far exceeds data traffic.

Telephone traffic is counted twice in this analysis since the bit rate is multiplied by two to account for the two-way nature of telephone calls. But email traffic is likewise multiplied by two since it counts emails made and received.

However, one service is missing in these estimates—fax communication—because my students rarely use it. But businesses are heavy users of faxes, and some telecom analysts claim that much of the traffic over international circuits is fax. For the sake of completeness in my estimates, assume that a typical manager sends an average of four fax pages per day. A high-resolution fax page is about 400,000b, assuming compression by a factor of 10. Thus, four pages of fax traffic would generate about 163105b—still much less than voice telephone traffic.

My estimates also do not include data traffic generated by the computer-to-computer communication of large businesses, so, for example, financial transactions are not included. But these types of data communication are probably minuscule because they contain very little data. Consider that on a very busy day, one billion shares of stock are exchanged on the New York Stock Exchange. Assuming 100b/share, the daily stock-market data traffic is 1003109b, or 1011b. This is a high estimate, since most transactions trade in blocks of 100 shares. Dividing by the 272 million people in the U.S. today gives a daily average per person of only 300b.

Telephone speech can be compressed to a bit rate of less than 64,000b/sec, and hence the estimates of telephone traffic might be too high. For example, digital wireless operates at a rate of about 12,000b/sec. But if telephone speech is compressed, email and Internet traffic can also be compressed. The ASCII code used for text requires 8b/character, even though the information rate for text is less than 2b/character.

It thus seems that the compression ratios for speech and for data are about the same, and my conclusions are unaffected.

T1-1Table 1. Summary of Traffic Estimates

T1-2Table 2. Details of Estimates

©1999 ACM  0002-0782/99/0600  $5.00

Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

The Digital Library is published by the Association for Computing Machinery. Copyright © 1999 ACM, Inc.

 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account
Article Contents: