Adding service discrimination to the Internet

Telecommunications Policy

Volumn 20, Issue 3 SPEC. ISS., 1996, Pages 169-181

  Clark, David D ^a  

^a MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

BANDWIDTH; DATA COMMUNICATION SYSTEMS; FREQUENCY ALLOCATION; RESOURCE ALLOCATION; TELECOMMUNICATION SERVICES; TELECOMMUNICATION TRAFFIC;

BANDWIDTH ALLOCATION; INTERNET;

COMPUTER NETWORKS;

EID: 0030125172     PISSN: 03085961     EISSN: None     Source Type: Journal    
DOI: 10.1016/0308-5961(96)00001-8     Document Type: Article

References (19)

1. Jacobson, V 'Congestion avoidance and control', Proceedings of ACM Sigcomm 1988 15 (4) 314-329
2. Another fact worth noting about the Internet is the rather long delays in the control mechanisms. With the cross-country round-trip delay of about 0.1 s, the response time of TCP to congestion information is necessarily slow. In this sort of time, many packets can be sent from a single source, and substantial short-term congestion can occur, which must be dealt with by queuing.
3. MacKie-Mason, J and Varian, H 'Pricing the Internet' presented at Public Access to the Internet, JFK School of Government, May 1993. Latest version available via ftp://ftp.econ.lsa.umich.edu/pub/Papers/ Pricing_the_Internet.ps.Z.
4. The typical round trip for a cross-country Internet path is about 0.1 s. The transmission time of a 2 kb data element on today's long-distance trunks (45 mb/s) is about 0.00035 s. Thus, even a queuing delay of 100 packets, adding 0.035 s of delay, is substantially less than the irreducible round-trip delay.
5. If the user fetches a new Web page on the average every 10 s, and each page on the average is 2 kb, then the data from each user occupies the trunk for 0.0035 s every 10 s. 28 000 of this sort of user can be supported in each direction across a 45 mb/s trunk. It is also worth noting that one such user generates data at about 2.5% of the rate of a single phone call. This relative efficiency is one of the characteristics of data transfers over the Internet.
6. Willinger, W, Murad, M, Sherman, R and Wilson, D 'Self-similarity through high-variability: statistica analysis of ethernet LAN traffic at the source level' Proceedings of ACM SigComm 1995 25 (4) 101-113
7. This adaptivity is also what allows TCP to transfer data over links of widely varying speeds, from dialup modems to 100 mb/s LANs and beyond. The current record for a long-distance TCP transfer is over 500 mb/s.
8. The Frame Relay service is defined in this way. The subscriber to a Frame Relay network must purchase a Permanent Virtual Circuit (PVC) between each source and destination for which a direct connection is desired. For each PVC, it is possible to specify a Committed Information Rate (CIR), which is the worst case rate that that PVC. Presumably, the provider must provision the network so that there are sufficient resources to support all the CIRs of all the clients. However, capacity not being used can be shifted at each instant to other users, so that the best case peak rate can exceed the CIR. This makes the service more attractive.
9. Demers, A, Keshav, S, and Shenker, S 'Analysis and simulation of a fair queuing algorithm' Proceedings of ACM SigComm 1992 19 (4) 1-12; also Demers, A, Keshav, S, and Shenker, S Research and Experience Journal of Internetworking 1990 1 3-26
10. Demers, A, Keshav, S, and Shenker, S 'Analysis and simulation of a fair queuing algorithm' Proceedings of ACM SigComm 1992 19 (4) 1-12; also Demers, A, Keshav, S, and Shenker, S Research and Experience Journal of Internetworking 1990 1 3-26
11. Clark, David D, Shenker, S, and Zhang, L 'Supporting real-time applications in an integrated services packet network: architecture and mechanism' Proceedings of ACM SigComm 1992 22 14-26
12. MacKie-Mason, J and Varian, H 'Pricing the Internet', presented at Public Access to the Internet, JFK School of Government, May 1993. Latest version available via ftp://ftp.econ.lsa.umich.edu/pub/Papers/ Pricing_the_Internet.ps.Z.
13. There is another drawback to the smart market scheme, which is that it couples the service model and the price model in a very direct way, which reduces the flexibility that providers have in setting price for the service.
14. Gupta, A, Stahl, D, and Whinston, A 'Managing the Internet as an economic system' available via http://cison. bus.utexas.ed/ravi/pricing.ps.Z
15. If there is no congestion, then there is presumably no queue of packets, which means that there is not a set of packets of different priority in the queue to reorder. Thus, priority scheduling normally has an effect only during congestion.
16. Clark, D 'A model for cost allocation and pricing in the Internet' in McKnight, L and Bailey, J (eds) Internet Economics Journal of Electronic Publishing, University of Michigan Press (forthcoming). Available via http://www.press.umich.edu:80/jep/
17. MacKie-Mason, J and Varian, H 'Pricing the Internet' presented at Public Access to the Internet, JFK School of Government, May 1993. Latest version available via ftp://ftp.econ.lsa.umich.edu/pub/Papers/ Pricing_the_Internet.ps.Z.
18. Of course, in the future, this whole payment pattern could be inverted. The long-distance providers could directly attract the individual subscribers, and then contract with the regional or local area providers to carry their traffic. This pattern now applies when businesses directly purchase long-distance telephone service from a provider, who then contracts with a CAP to connect the business to that provider.
19. Shenker, S, Clark, D, Estrin, D, and Zhang, L 'Pricing in computer networks: reshaping the research agenda' this issue of Telecommunications Policy