Distributed QoS Control: Pricing in the Internet


Distributed control of Quality of Service (QoS) is a critical issue in emerging networks. There is now a substantial body of work on traffic characterization and application requirements characterization, which will allow users to describe the QoS requirements of their traffic flows. There is also a substantial body of work on integrated and differentiated services network architectures, which will allow networks to give differentiated performance to these flows.

However, without a distributed control model to manage QoS within the network, the network can not efficiently transform such user characterizations into network resource allocations and make intelligent decisions as to how to treat each traffic flow. In an integrated services architecture such as Internet's IntServ or ATM's Virtual Path model, some mechanism needs to decide what network resources to reserve for each traffic flow, or for aggregated traffic flows. Similarly, in a differentiated service architecture such as Internet's DiffServ, some mechanism needs to decide what priority each packet should receive, i.e. how to assign codepoints.

In commercial networks, QoS specification will likely be based on application requirements and traffic characteristics. In a military network, however, QoS may also depend on the mission that the traffic supports. Some measure of the relation of communication to the success of a mission may be useful in determining what QoS that flow should receive when the network is congested.

In both ATM and Internet, therefore, some connection needs to be made between the user and networks sides of the QoS contract. Given user characterizations for heterogeneous services, including both real-time and non real-time, with different definitions of QoS, and given some characterization of network availability, how should it be decided which users are granted service? How much buffer and bandwidth should be allocated to each service, if not enough are free to serve all users? How should allocation of bandwidth and buffer take into consideration various QoS? What is the tradeoff between competing services with differing QoS, especially between real-time and non real-time services? It is essential that these questions be answered to bridge the gap between the user and network sides of the QoS contract.

In this project, we focus on a distributed QoS control model. We are concerned with how networks resources (e.g. buffer, bandwidth, power, codes) should be allocated, based on information gathered from both the user and the network. Distributed QoS control mechanisms will interface with traffic and network models through traffic characterizations, application performance characterizations, and network resource management architectures. In a simulation environment, the control model will interact with user and network modules using discrete event signals. In the network itself, the QoS control mechanism would specify what information is exchanged among users and network components in order to efficiently allocate network resources.

The resulting control model will be distributed, and will minimize collection of user information and network overhead. Furthermore, it will automate resource management and QoS management tasks that are currently accomplished through human intervention, if at all. The control model will apply to individual traffic flows and to aggregations of traffic flows, so that it is scalable with network size and capacity. It will apply to resource dynamics that are on a time scale of a round trip time or larger. Finally, the proposed control model will provide information (shadow costs) about the value of various network capacity increases, so that network dimensioning becomes more automated.

Principal Papers:

Connection Establishment in High Speed Networks (with H. Jiang), IEEE Journal on Selected Areas in Communications, vol. 13 no. 7, September 1995, pp. 1150-1161.

The Role of Price in the Connection Establishment Process (with H. Jiang), European Transactions on Telecommunications, vol. 6 no. 4, July-August 1995, pp. 421-429. (journal, paper)

The Effect of Bandwidth and Buffer Pricing on Resource Allocation and QoS (with N. Jin), Computer Networks, Special Issue on Internet Economics: Pricing and Policies, vol. 46 no. 1, September 2004, pp. 53-71.

Dynamic Congestion-Based Pricing of Bandwidth and Buffer (with N. Jin and G. Venkitachalam), IEEE/ACM Transactions on Networking, December 2005.

Information Exchange in diffServ Pricing (with N. Jin), IEEE Globecom, November 2005. (abstract, paper)

Additional Papers:

A Pricing Model for High Speed Networks with Guaranteed Quality of Service (with H. Jiang), IEEE InfoCom, March 1996, pp. 888-895. (abstract, paper)

A Pricing Model for Networks with Priorities (with H. Jiang and I. Sidhu), Allerton Conference on Communication, Control and Computing, September 1996, pp. 269-275. (conference site, paper)

Pricing of Buffer and Bandwidth in a Reservation-Based QoS Architecture, IEEE International Conference on Communications, Ancourage, Alaska, May 2003, pp.1521-1525. (abstract, paper)

Characteristics of Resource Allocation using Pricing (with N. Jin and G. Venkitachalam), IEEE Computer Communications Workshop, Laguna Niguel, California, October 2003, pp. 59-65. (abstract, paper)

Dynamic Pricing of Network Resources (with N. Jin and G. Venkitachalam), IEEE Globecom, San Francisco, California, December 2003, pp. 3216-3220. (abstract, paper)

Sensitivity of Optimal Quality of Service to Bandwidth and Buffer Prices (with N. Jin), IEEE Conference on Decision and Control, Maui, Hawaii, December 2003, pp. 1580-1585. (abstract, paper)

Portions of this work were supported by DARPA and NSF. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation, DARPA, or IEEE. This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. One print or electronic copy may be made for personal use only. Permission must be obtained from the copyright holder for systematic or multiple reproduction, distribution to multiple locations via electronic or other means, duplication of any material in these papers for a fee or for commercial purposes, modification of the content of these papers, reprinting or republishing of this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, and to reuse any copyrighted component of this work in other works.


Scott Jordan (6/24/2005)