Computer Networks and ISDN Systems
A quality of service negotiation approach with future reservations (NAFUR): a detailed study
Introduction
The new distributed multimedia (MM) applications are characterized by handling continuous media and by managing various media at the same time. Different types of continuous media require different levels of quality of service (QoS), and they require guarantees for the level of service to be maintained. This implies stringent requirements for the communication systems and the end-systems to support the requirements of MM applications. Hence, these applications need end-to-end QoS management, particularly QoS negotiation, to ensure that the requirements of the users are satisfied.
Most existing QoS negotiation protocols 1, 2, 3, 4, 5, 6, 7, 8are only concerned with the communication quality in terms of QoS parameters, such as throughput, delay and jitter. Furthermore the negotiation results, in response to the user request, are restricted to an acceptance or rejection of the request. This implies that a second attempt of the user cannot take advantage of information obtained through the first request to change, if possible, the requirements to fit the current system load. In [9]a QoS negotiation protocol based on the Tenet protocol suite [10]has been proposed to improve the information given to the users by the network when a connection request is rejected. Such improvement is closely related to the characteristics of the Tenet protocol suite, e.g. admission control tests. Also application-to-application negotiation protocols 11, 12, 13, 14focus only on the establishment of an agreement between the parties with respect to the application QoS parameters. More generally, the service model of the existing negotiation approaches provides the user with the QoS that can be supported at the time the service request is made, and assumes that the service is requested for indefinite duration. We believe that such approaches do not fit the needs for future MM service providers and users. Let us present an example which motivates this claim.
We consider a video-on-demand application which supports remote access to MM databases (Fig. 1). In the following we present two basic situations of user–server interactions in the existing negotiation approaches.
(a) Let us assume that a user located at client-1 asks to play a movie with a desired QoS, e.g. (video_color=color, video_rate=TV-rate, video_window_size=large), located at server-1. To support this service request server-1, MM transport system and client-1 must commit to reserve resources to support a level of QoS, e.g. delay, in such way that the end-to-end QoS is satisfied [15]. Unfortunately, at the time the request is made there are not enough resources to support the service requested. Making use of the existing negotiation approaches a rejection will be sent to the user.
A desirable QoS negotiation approach will provide the user with two proposals: (1) the requested service can be provided immediately with a degraded QoS, e.g. (video_color=black and white, video_rate=TV-rate, video_window_size=large), and (2) the requested service can be provided with the desired QoS at a future time T1, e.g. 30 minutes later. Thus, the user can choose between the two proposals depending on his/her requirements.
(b) Let us assume that the user selects the second proposal, and at a later time T2<T1, a second user on client-2 asks for the same movie. Fortunately, the system has enough resources to support the second user request at T2 (immediately). Making use of the current negotiation approaches an acceptance will be sent to the user.
A desirable negotiation approach will check the possibility of supporting the second user request concurrently with the first user at time T1, since this may lead to sharing of resources and possibly lower cost. If this is the case, two proposals will be sent to the second user: (1) an acceptance to start the movie immediately (at T2) with a cost cost1, and (2) an acceptance to start the movie later at T1 with a cost cost2 where cost2<cost1. Consequently, the user can select the proposal which corresponds to his/her wishes and constraints. The computation of the second proposal is motivated by the fact that servicing users individually is inefficient, expensive and not scalable. One of the best ways to deliver information to more users is through the use of multicast communication which has the advantage of being scalable. We believe that a large number of users will select the second proposal which will optimize resources usage, e.g. multicast communication [16], and minimize cost for the users.
To increase the flexibility and the availability of future MM systems (a) the starting time of the requested service should be decoupled from the time the service request is made, and (b) the duration of the service requested should be specified by the service user. In the present paper we propose a QoS negotiation approach with future reservations (NAFUR) that supports these principles. To make use of NAFUR, the user must specify, besides the desired QoS, only the duration of the requested service.
There is a basic assumption in our approach: We assume that the system is built in the framework of QoS guarantees, that is, the components are able to reserve resources to support certain levels of QoS. Consequently, NAFUR provides the flexibility to incorporate a range of existing resource reservation schemes and scheduling policies, and a range of new system component technologies, such as ATM.
The paper is organized as follows. Section 2describes the operation of NAFUR at the user interface; it also presents some performance analysis of NAFUR by means of simulations. Section 3describes the operation of NAFUR in a hierarchical multi-domain environment; some ideas to optimize the operation of NAFUR are also discussed. Section 4discusses the case of unknown services durations. Finally, Section 5summarizes our results and presents some concluding remarks.
Section snippets
QoS negotiation interactions at the user interface for present and future service sessions
In this section we model a distributed system as a single module and a QoS manager (Fig. 2) which represents the access point to the module. The ability of the QoS manager to process a service request is directly related to the admission criteria which the QoS manager uses to decide whether a new request is accepted. This criteria is that the sum of previously assigned resources plus the resources required by the new request should not exceed the resources of the system. In the case of
A simple hierarchical negotiation of QoS with present and future service sessions
In Section 2, we assumed that a distributed system consists of a single module and a QoS manager (Fig. 2); this assumption is not realistic since a distributed system consists of a number of components, e.g. networks and servers. That is, a set of components are required to provide the QoS associated with a requested service. For example, in the context of a news-on-demand service, data is read from the disk, stored in buffers at the sender side, transmitted over the network and again stored in
The case of unknown service durations
For a range of multimedia applications, it is difficult to predict the duration of a session. For example, collaborative systems assume that discussions are somewhat unstructured and therefore it is almost impossible to know, in advance, the (accurate) duration of a collaborative session. In such a situation, the reservation algorithms described above cannot be used directly, since they assume that the service duration is known when the service request is made. We discuss in the following how
Conclusion
This paper describes work on a new QoS negotiation approach with future reservations (NAFUR) that decouples the starting time of the requested service from the time the service request is made. It is assumed that the duration of the requested service requested is known. NAFUR allows to compute the QoS that can be supported at the time the service request is made, and at certain later times carefully chosen. For example, if the requested QoS cannot be supported at the time the service request is
Acknowledgements
This work was supported by a grant from the Canadian Institute for Telecommunication Research (CITR), under the Networks of Centres of Excellence Program of the Canadian Government.
Abdelhakim Hafid is Assistant Professor at the Electrical & Computer Enggineering/Compter Science Departments (a joint appointment), University of Western Ontario, and a Research Director of the Advanced Communication Engineering Centre (venture established by UWO, Bay Networks, Bell Canada); he is also an Adjunct Professor at the University of Montreal, Department of Computer Science. He received his Masters and Ph.D. degrees in computer science from the University of Montreal on quality of
References (23)
- D. Anderson, R. Herrtwich, C. Schaefer, SRP: a resource reservation protocol for guaranteed-performance communications...
- C. Chou, K. Shin, Statistical real-time video channels over a multi-access network, Proc. High SPIE,...
- A. Danthine, OSI95: high performance protocol with multimedia support on HSLANs and B-ISDN, 3rd Joint European...
- D. Ferrari, D. Verma, A scheme for real-time channel establishment in wide-area networks, IEEE JSAC 8 (3)...
- D. Kandlur, K. Shin, D. Ferrari, Real-time communication in multi-hop networks, Proc. 11th Int. Conf. on Distributed...
- B. Metzler, I. Miloucheva, K. Rebensburg, Multimedia communication platform: specification of the broadband transport...
- C. Topolcic, Experimental Internet stream protocol: version 2 (ST-II), Internet RFC 1190,...
- D. Hehmann, R. Herrtwich, W. Schulz, T. Schett, R. Steinmetz, Implementing HeiTS: architecture and implementation...
- J. Ramaekers, G. Ventre, Quality of service negotiation in a real-time communication network, Technical Report...
- D. Ferrari, A. Banerjea, H. Zhang, Network support for multimedia, Technical Report 92-072, International Computer...
Cited by (22)
End-to-end quality of service for high-end applications
2004, Computer CommunicationsCitation Excerpt :However, there has been little work on the specific problems addressed in this paper, namely, advance reservation and co-reservation of heterogeneous collections of resources for end-to-end QoS and the use of DS mechanisms to support flow types encountered in high-end applications. Proposals for advance reservations typically employ cooperating servers that coordinate advance reservations along an end-to-end path [14,17,27,52]. Techniques have been proposed for representing advance reservations, for balancing immediate and advance reservations [17], and for advance reservation of predictive flows [14].
A scalable video-on-demand system using future reservation of resources and multicast communications
1998, Computer CommunicationsGrid resource management: On-demand provisioning, advance reservation, and capacity planning of grid resources
2010, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)A predictive admission control algorithm for user advance reservation in equipment grid
2009, International Journal of Business Process Integration and ManagementAdvanced reservation in Grid using PBS
2008, Proceedings of the Workshop on Enabling Technologies: Infrastructure for Collaborative Enterprises, WETICEPredictive admission control algorithm for advance reservation in equipment grid
2008, Proceedings - 2008 IEEE International Conference on Services Computing, SCC 2008
Abdelhakim Hafid is Assistant Professor at the Electrical & Computer Enggineering/Compter Science Departments (a joint appointment), University of Western Ontario, and a Research Director of the Advanced Communication Engineering Centre (venture established by UWO, Bay Networks, Bell Canada); he is also an Adjunct Professor at the University of Montreal, Department of Computer Science. He received his Masters and Ph.D. degrees in computer science from the University of Montreal on quality of service management for distributed multimedia applications in 1993 and 1996, respectively. From 1996 to 1997 he was a Researcher Staff Member at the Computer Research Institute of Montreal (CRIM), Telecommunications and Distributed Systems Division, working in the area of distributed multimedia applications. From 1993 to 1994 he was visiting scientist at GMD-FOKUS, Systems Engineering and Methods group, Berlin, Germany working in the area of high speed protocols testing. His current research interests are in Internet and multimedia networking.
Gregor von Bochmann is a professor at the University of Montreal since 1972 and holds the Hewlett-Packard-NSERC-CITI chair of industrial research on communication protocols. He is also one of the scientific directors of the Centre de Recherche Informatique de Montréal (CRIM), and a Fellow of the IEEE and ACM. Professor von Bochmann has worked in the areas of programming languages, compiler design, communication protocols, and software engineering and has published many papers and some books in these areas. He has also been actively involved in the standardization of formal description techniques for OSI communication protocols and services. From 1977 to 1978 he was a visiting professor at the Ecole Polytechnique Fédérale, Lausanne, Switzerland. From 1979 to 1980 he was a visiting professor in the Computer Systems Laboratory, Stanford University, California. From 1986 to 1987 he was a visiting researcher at Siemens, Munich, Germany. His present work is aimed at methodologies for the design, implementation and testing of communication protocols and distributed systems. Ongoing projects include applications to high-speed protocols, distributed systems management and quality of service negotiation for distributed multimedia applications.
Rachida Dssouli is professor in the Département d'Informatique et de Recherche Opérationnelle (DIRO), Université de Montréal. She received the Doctorat d'Université degree in computer science from the Université Paul-Sabatier of Toulouse, France, in 1981, and a Ph.D. degree in computer science in 1987, from the University of Montreal. Professor Dssouli has been professor at the Université Mohamed 1er, Oujda, Morocco, from 1981 to 1989, and assistant professor at the Université de Sherbrooke, from 1989 to 1991. She is currently on Sabbatical at NORTEL, Ile des Soeurs. Her research area is protocol engineering and requirements engineering. Ongoing projects include incremental specification and analysis of reactive systems based on scenario language, multimedia applications and tests of timed communicating systems.