Pricing in computer networks: Reshaping the research agenda

Telecommunications Policy

Volumn 20, Issue 3 SPEC. ISS., 1996, Pages 183-201

  Shenker, Scott ^a   Clark, David ^b   Estrin, Deborah ^c   Herzog, Shai ^c  

^a PALO ALTO RESEARCH CENTER   (United States)

^b MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

^c UNIVERSITY OF SOUTHERN CALIFORNIA   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

COMPUTER ARCHITECTURE; COSTS; MATHEMATICAL MODELS; PUBLIC POLICY; TELECOMMUNICATION SERVICES;

FLAT PRICING; INTERNET; USAGE BASED CHARGING;

COMPUTER NETWORKS;

EID: 0030121216     PISSN: 03085961     EISSN: None     Source Type: Journal    
DOI: 10.1016/0308-5961(96)00002-X     Document Type: Article

References (60)

1. This paper is also being published in the proceedings of the Twenty-Third Annual Telecommunications Policy Research Conference, edited by Gerald Brock and Greg Rosston, and published by Lawrence Erlbaum Associates. This research was supported in part by the Advanced Research Projects Agency, monitored by Fort Huachuca under contracts DABT63-94-C-0073 (SS), DABT63-94-C-0072 (DC), and DABT63-91-C-0001 (DE, SH). The views expressed here do not reflect the position or policy of the US government.
2. See, for instance, 'Traffic jams already on the information highway' New York Times 3 November 1993 A1; 'Planet Internet' Business Week April 3 1995
3. See, for instance, 'Traffic jams already on the information highway' New York Times 3 November 1993 A1; 'Planet Internet' Business Week April 3 1995
4. The authors include themselves in this critique, having adhered to this optimality paradigm in previous publications, such as Cocchi, R, Estrin, D, Shenker, S and Zhang, L 'A study of priority pricing in multiple service class networks' Proceedings of SigComm '91 September 1991 123-130; and Cocchi, R, Estrin, D, Shenker, S and Zhang, L 'Pricing in computer networks: motivation, formulation, and example' ACM/IEEE Transactions on Networking 1993 1 614-627
5. The authors include themselves in this critique, having adhered to this optimality paradigm in previous publications, such as Cocchi, R, Estrin, D, Shenker, S and Zhang, L 'A study of priority pricing in multiple service class networks' Proceedings of SigComm '91 September 1991 123-130; and Cocchi, R, Estrin, D, Shenker, S and Zhang, L 'Pricing in computer networks: motivation, formulation, and example' ACM/IEEE Transactions on Networking 1993 1 614-627
6. We use the term stable only to mean that the revenues cover costs; we do not use the term to refer to any other dynamical properties of the equilibrium.
7. MacKie-Mason, J and Varian, H 'Pricing the Internet' in Kahin, B and Keller, J (eds) Public Access to the Internet Prentice-Hall, Englewood Cliffs (1995) 269-314
8. The seminal references on this phenomenon are: Strotz, R H 'Urban transportation parables' in Margolis, J (ed) The Public Economy of Urban Communities Resources for the Future, Washington, DC (1965) 127-169; and Mohring, H and Hartwitz, M Highway Benefits: An Analytical Approach Northwestern University Press, Evanston (1962)
9. The seminal references on this phenomenon are: Strotz, R H 'Urban transportation parables' in Margolis, J (ed) The Public Economy of Urban Communities Resources for the Future, Washington, DC (1965) 127-169; and Mohring, H and Hartwitz, M Highway Benefits: An Analytical Approach Northwestern University Press, Evanston (1962)
10. For a more thorough discussion of this point, see Srinagesh, P 'Internet cost structure and interconnection agreements' technical report. Bellcore (1995);
11. Gon, J and Srinagesh, P 'The economics of layered networks' technical report, Bell Communications Research, Inc (1995)
12. We ignore non-linear pricing policies here in order to simplify the discussion.
13. We should make clear that we are assuming that network providers are not also controlling, or directly profiting from, the content delivered over their networks. However, in bundled networks such as cable TV, where the application and the network transport are sold as a single entity, there are many more opportunities to recover costs. Profits on content and revenue from advertising are important aspects of pricing in bundled networks. We do not consider such bundled networks in this paper but restrict ourselves to the analysis of pricing pure Internet access without bundled services. The nature of the ISP market is still very much in flux, and there may be other sources of revenue in the future, such as renting space on provider-supplied Web-servers, that may complicate the rather simplified case we are analyzing.
14. It is possible that, in the grand and glorious future, the GII will have a single Internet-like network infrastructure and all households will have a single network connection that carries their telephony, television and data traffic. At that point, it may well be true that essentially all users have high valuation of their network connection and raising attachment fees is the appropriate way to raise revenue. However, we are a long way from this Utopian vision, and we should design our current network pricing policies to fit the present situation.
15. This assumes seamless interconnection, so switching providers does not affect connectivity. Otherwise, the decision to switch providers involves many other factors besides cost.
16. As in, for example, Shenker, S 'Making greed work in networks: a game-theoretic analysis of switch service disciplines' Proceedings of SigComm '94 August 1994 47-57
17. See the discussion about the properties of best-effort traffic in Clark, D 'A model for cost allocation and pricing in the Internet' technical report, MIT (August 1995)
18. See Clark op cit Ref 13 for a more thorough discussion of this point.
19. The implications of this layering for content provision is discussed in Mackie-Mason, J, Shenker, S and Varian, H 'Service architecture and content provision' technical report, University of Michigan and Xerox PARC (1995)
20. A few representative samples of the pricing literature are: Edell, R, McKeown, N and Varaiya, P 'Billing users and pricing for TCP' IEEE Journal on Selected Areas in Communications 1988 13 1162-1175; McLean, R and Sharkey, W 'An approach to the pricing of broadband telecommunications services' technical report, Bellcore (March 1993); Stahl, D and Whinston, A 'An economic approach to client- server computing with priority classes' technical report, University of Texas at Austin (1992)
21. A few representative samples of the pricing literature are: Edell, R, McKeown, N and Varaiya, P 'Billing users and pricing for TCP' IEEE Journal on Selected Areas in Communications 1988 13 1162-1175; McLean, R and Sharkey, W 'An approach to the pricing of broadband telecommunications services' technical report, Bellcore (March 1993); Stahl, D and Whinston, A 'An economic approach to client- server computing with priority classes' technical report, University of Texas at Austin (1992)
22. A few representative samples of the pricing literature are: Edell, R, McKeown, N and Varaiya, P 'Billing users and pricing for TCP' IEEE Journal on Selected Areas in Communications 1988 13 1162-1175; McLean, R and Sharkey, W 'An approach to the pricing of broadband telecommunications services' technical report, Bellcore (March 1993); Stahl, D and Whinston, A 'An economic approach to client-server computing with priority classes' technical report, University of Texas at Austin (1992)
23. MacKie-Mason, J and Varian, H 'Pricing the Internet' in Kahin, B and Keller, J (eds) Public Access to the Internet Prentice-Hall, Englewood Cliff (1995);
24. MacKie-Mason, J, Murphy, J and Murphy, L 'ATM efficiency under various pricing schemes' technical report, University of Michigan, Dublin City University and University of Auburn (March 1995);
25. MacKie-Mason, J, Murphy, J and Murphy, L 'The role of responsive pricing in the Internet' technical report, University of Michigan, Dublin City University and University of Auburn (June 1995)
26. As an aside, note that the pricing scheme is embedded within the architecture in this proposal. The bids are translated into the prices charged.
27. If one believes that the major source of congestion is at the edge of the network, then one could apply the smart market only at the edge points. This removes the end-to-end versus per-hop problem and could be used in our edge pricing scheme as the method of charging. See the discussion in the section titled 'Forms of pricing'.
28. Wang, Q, Sirbu, M and Peha, J 'An optimal pricing model for cell-switching integrated services networks' technical report, Carnegie Mellon University (May 1995)
29. Gupta, A, Stahl, D and Whinston, A 'Managing the Internet as an economic system' technical report, University of Texas at Austin (July 1994)
30. Gupta, A, Stahl, D and Whinston, A 'The Internet: a future tragedy of the commons' technical report, University of Texas at Austin (1995); MacKie-Mason, J and Varian, H 'Pricing congestible network resources' IEEE Journal on Selected Areas in Communication 1995 13 1141-1149; MacKie-Mason, J, Murphy, J and Murphy, L 'The role of responsive pricing in the Internet' technical report, University of Michigan, Dublin City University and University of Auburn (June 1995)
31. Gupta, A, Stahl, D and Whinston, A 'The Internet: a future tragedy of the commons' technical report, University of Texas at Austin (1995); MacKie-Mason, J and Varian, H 'Pricing congestible network resources' IEEE Journal on Selected Areas in Communication 1995 13 1141-1149; MacKie-Mason, J, Murphy, J and Murphy, L 'The role of responsive pricing in the Internet' technical report, University of Michigan, Dublin City University and University of Auburn (June 1995)
32. Gupta, A, Stahl, D and Whinston, A 'The Internet: a future tragedy of the commons' technical report, University of Texas at Austin (1995); MacKie-Mason, J and Varian, H 'Pricing congestible network resources' IEEE Journal on Selected Areas in Communication 1995 13 1141-1149; MacKie-Mason, J, Murphy, J and Murphy, L 'The role of responsive pricing in the Internet' technical report, University of Michigan, Dublin City University and University of Auburn (June 1995)
33. The ability to assign charges to the receiving end could, in some cases, be handled by a higher level protocol that redistributes the basic charges determined by the network. However, there are several disadvantages to requiring such a high level protocol: it requires the ability to transfer funds at a higher level, it cannot deal with capacity-based charging, and in the multicast case the required information (such as the membership of the group and the network topology) may not be available at the higher layer. Thus, we think it preferable to build the flexibility of assignment into the basic charging mechanism itself.
34. Srinagesh op cit Ref 7; Gon and Srinagesh op cit Ref 7
35. Srinagesh op cit Ref 7; Gon and Srinagesh op cit Ref 7
36. See MacKie-Mason, Murphy, and Murphy op cit Ref 22 for a similar discussion of the role of adaptation.
37. Jacobson, V, private communication (1995)
38. Clark op cit Ref 13
39. We use the term 'bill' here only to connote that the packet is applied to the contract the user has with provider A; as we mention below, the contracted pricing policy may very well be a flat price with a limit on peak rate, in which case there is no additional charge per packet.
40. The distinction between fixed and variable prices may be extremely important to individual users; users on fixed budgets may need fixed prices, whereas users with extremely variable demand may need the ability to only pay for usage. Our point is that this distinction, while important to individual users, is not fundamentally important from an architectural or economic perspective. Both forms of pricing can be assessed locally, and both constrain usage.
41. In our taxonomy, attachment prices would refer only to the price of attaching to the network and not refer at all to the speed of the access line. All other charges would be considered usage-constraining. Of course, in non-linear pricing schemes (or when there is a spectrum of pricing menus offered, as in the current cellular telephony market) the distinction between the two is completely blurred.
42. There may be some disadvantages with using the smart-market as the local pricing scheme (eg it embeds the pricing policy in the architecture), but our point here is that it is not architecturally precluded by the edge pricing paradigm, and so firms are free to experiment with it.
43. Charging receivers for service is a non-local problem because, in approaches with explicit willingness-to-pay signaling, when both the source and receiver are serviced by the same provider the source's access point must be informed that the receiver is willing to assume responsibility for the transmission. Similarly, when the path from source to receiver traverses several different provider networks, the notification of receiver-paying must be communicated to both the exit access point and the entrance access point in each network. Other approaches can avoid this explicit nonlocal signaling by adopting some uniform standards, such as a certain portion of the multicast address space being set aside for receiver-pay groups, but these standards themselves are nonlocal in that they represent agreements between providers about a billing policy.
44. In particular, there is a spectrum of design choices providing different levels of functionality and requiring more or less additional mechanism; in the following discussion we are not attempting to make a detailed evaluation of the functionality versus mechanism tradeoff, but are merely illustrating some possible ways of achieving the aforementioned goals. There are more minimalist approaches to these problems that require less additional mechanism, and they should be considered when making design decisions for the Internet, but for this pedagogical discussion we have presented more straightforward, if more mechanistic, approaches.
45. These tables would contain information describing how many usage units (for a capacity filter) each packet represents, or a monetary per-packet charge, or whatever other information is needed for the provider to assess the appropriate charges.
46. Deering, S, private communication (1994)
47. Deering, S and Hinden, R 'Internet protocol, version 6 (ipv6) specification' Internet draft (June 1995)
48. For more information on addressing options, see Francis, P 'Comparison of geographical and provider-rooted Internet addressing' Proceedings of INET '94 June 1994
49. One could apply such a scheme to a logically overlaid network so the prices would be less dependent on the details of the path. For instance, the network could be divided up into area codes, with logical link costs recorded whenever the accounting message left one area code and entered another.
50. Zhang, L, Deering, S, Estrin, D, Shenker, S and Zappala, D 'RSVP: a resource reservation protocol' IEEE Network Magazine 1993 7 8-18
51. Herzog, S, Shenker, S and Estrin, D 'Sharing the "cost" of multicast trees: an axiomatic analysis' Proceedings of Sig-Comm '95 August 1995 315-327
52. Gawlick, R 'Admission control and routing: theory and practice' technical report, MIT (August 1995)
53. Herzog et al op cit Ref 40
54. The seminal paper on the topic is: Deering, S and Cheriton, D 'Multicast routing in datagram internetworks and extended LANs' ACM Transactions on Computer Systems 1990 8 (2) 85-110
55. Casner, S and Deering, S 'First IETF Internet audiocast' Computer Communications Review 1992 22 92-97
56. Shenker, S 'Some fundamental design decisions for the future Internet' IEEE Journal on Selected Areas in Communications 1995 13 1176-1188
57. Clark op cit Ref 13
58. These issues are discussed in the references cited in Ref 3, and also in: Clark, D, Zhang, L and Shenker, S 'Supporting real-time applications in an integrated services packet network: architecture and mechanism' Proceedings of Sigcomm '92 ACM Press, New York (1992) 14-26; Shenker, S 'Service models and pricing policies for an integrated services Internet' in Kahin, B and Keller, J (eds) Public Access to the Internet Prentice-Hall, Englewood Cliff (1995) 315-337
59. These issues are discussed in the references cited in Ref 3, and also in: Clark, D, Zhang, L and Shenker, S 'Supporting real- time applications in an integrated services packet network: architecture and mechanism' Proceedings of Sigcomm '92 ACM Press, New York (1992) 14-26; Shenker, S 'Service models and pricing policies for an integrated services Internet' in Kahin, B and Keller, J (eds) Public Access to the Internet Prentice-Hall, Englewood Cliff (1995) 315-337
60. A slightly out-of-date overview of this proposed architecture is presented in: Braden, R, Clark, D and Shenker, S 'Integrated services in the Internet architecture: an overview' RFC 1633 (June 1994)