Optimizing multimedia transcoding multicast trees
Introduction
The current internet is very heterogeneous: Users are connected to the net in different ways with a wide variety in access capabilities. Nevertheless they all want the best quality possible and in case of multimedia streams, each user expects to receive the stream at the best quality possible. If the provider sends out only one multicast stream, there are currently two possibilities: either a high quality stream is sent and consequently users with a slow connection will not be able to receive the stream or either a low quality stream is transmitted, resulting in displeased broadband access users.
A solution to this problem consists of sending multiple formats of the same stream towards the requesting users (simulcasting [1]). This implies that the provider is responsible for all necessary transcodings and should send out all different formats in separate multicast trees.
Another possibility is to use layered media streams and let the users subscribe to multiple layers to get better quality as proposed by McCanne in [2]. This method can only be used with formats supporting this layering of data.
The solution discussed in this paper consists of extending (some of) the network nodes with transcoding capabilities making them ‘active nodes’ ([3], [4], [5], [6]). In this way, only one multicast tree has to be set up and the transcodings can be done inside the network, at optimal locations i.e. at nodes that have enough processing power to do the transcoding and chosen such that the overall bandwidth usage is minimized. As a result, the provider only needs to inject a single format into the active (overlay) network while the users still get a customized service.
Fig. 1 illustrates this solution: a video multicast server sends a video stream at 6 Mbps into the network. Three users request the video: user A requests high quality and asks for the video at full rate. User B also has a cable modem but requires only a 3 Mbps signal. User C has a T1 connection and requests the video at 1.2 Mbps. The active solution presented in Fig. 1 shows that the 6 Mbps signal is injected in the tree and that in two nodes a transcoding takes place, thereby optimizing the overall bandwidth usage.
The problem at hand has been described by Parnes [7] and the ‘active’ approach has been studied in a.o. [8], [9], [1], [10], [11]. Pasquale et al. [8] use the concept of the ‘relocatable continuous media filter’: filters propagate upstream towards the source to decrease bandwidth consumption. In [9], a video gateway is used to perform bandwidth adaptation to match the transmission quality to the heterogeneous bandwidth constraints of distinct regions of a single logical multicast session. In [1], Kouvelas et al. propose a scheme that uses self-organization to form groups out of co-located receivers with bad reception and that provides local repair through the use of transcoders. In [10], a centralized algorithm is proposed that builds a hierarchy of multicast groups to tackle the problem. Singh et al. [11] propose the use of proxies that can cache and transcode Web content.
In the work proposed here we focus on finding the optimal transcoding locations, which has not been addressed yet. The Transcoding Multicast Tree problem calculates the minimum cost multicast tree from source to all destinations and it determines the optimal transcoding locations in this multicast tree. The cost is related to the bandwidth usage on the edges and the resource usage in the nodes. An optimal solution and two heuristics are presented. The optimal solution will be used as a reference for the heuristics.
The remainder of this paper is organized as follows. The ILP-formulation of the problem is given in Section 2. An optimal solution can be calculated using this formulation. In Section 3, two heuristics that solve the problem in a non-optimal but faster way are presented. In Section 4, simulation results to evaluate the proposed algorithms and the influence of different parameters (number of destinations, number of types, number of active nodes, etc.) are shown. Finally, the conclusions are given in Section 5.
Section snippets
ILP-formulation of the problem
In order to find an optimal solution to the Transcoding Multicast Tree (TMT) problem and also to present a precise view on the problem, an ILP-formulation is given here.
The problem under investigation deals with calculating a minimum cost tree, taking into account both bandwidth and node resources. It is clear that this is related to and is even harder than the Minimal Steiner Tree problem [13]. Since the Minimal Steiner Tree problem is NP-complete, it is obvious that our problem is also
Heuristics
As mentioned before, the TMT problem is NP complete, so finding the optimal solution will not be feasible for large networks or for a large number of types or destinations. Therefore, two heuristics are proposed here, offering a fast but sub-optimal solution.
Simulations
Several simulations were done to illustrate the usefulness of the active solution and to compare the performance of the different solutions discussed in Sections 2 ILP-formulation of the problem, 3 Heuristics.
We also implemented the simulcasting [1] strategy to compare the ‘active’ solution with a ‘non-active’ solution. Exact least cost multicast trees where calculated for implementing this method, yielding the optimal non-active solution. For this solution, the tree cost is the sum of the
Conclusions
In this paper, solutions for the transcoding multicast tree problem in an active network were presented. The optimal solution was found using the ILP-formulation of the problem. Two heuristics were proposed to solve the problem for realistic problem sizes. The Skinned Steiner Tree heuristic, tries to ‘skin’ a multicast tree with enough bandwidth in order to determine the bandwidth and transcodings needed. The Branch Attach heuristic starts with a multicast tree for the highest quality class and
Acknowledgements
This work was carried out within the framework of the project CoDiNet sponsored by the Flemish Institute for the promotion of Scientific and Technological Research in the Industry (IWT). It is in part supported by the Ghent University GOA-project ‘Programmable and Active Networks’.
Thijs Lambrecht received his Masters Degree in Computer Science (option Telecommunication Networks) in 1999 from the Ghent University, Belgium. He is now a Ph.D. student at the same University where he is part of the Department of Information Technology (INTEC). He works in the Integrated Broadband Communications Network group (IBCN). He worked on “Routing with revenue optimalization” as part of his graduate thesis. His current research interests include Active Networks where his focus is on
References (23)
- I. Kouvelas, V. Hardman, J. Crowcroft, Network adaptive continuous media applications through self organised...
- S. McCanne, Van Jacobson, M. Vetterli, Receiver-Driven Layered Multicast, SIGCOMM, Stanfort, CA, August 1996, pp....
A survey of active network research
IEEE Commun. Mag.
(1997)- et al.
Activating networks: a progress report
IEEE Computer Mag
(1999) - et al.
Directions in active networks
IEEE Commun. Mag.
(1998) - K. Psounis, Active networks: applications, security, safety, and architectures, IEEE Communications Surveys, First...
- P. Parnes, Media distribution in heterogeneous environments using IP-Multicast, 1998. Available from:...
- J.C. Pasquale, G.C. Polyzos, E.W. Anderson, V.P. Kompella, Filter propagation in dissemination trees: trading off...
- E. Amir, S. McCanne, H. Zhang, An Application Level Video Gateway, in: Proceedings of the ACM Multimedia ’95, San...
- H. Akamine, N. Wakamiya, M. Murata, H. Miyahara, On the construction of heterogeneous multicast distribution trees...
Cited by (0)
Thijs Lambrecht received his Masters Degree in Computer Science (option Telecommunication Networks) in 1999 from the Ghent University, Belgium. He is now a Ph.D. student at the same University where he is part of the Department of Information Technology (INTEC). He works in the Integrated Broadband Communications Network group (IBCN). He worked on “Routing with revenue optimalization” as part of his graduate thesis. His current research interests include Active Networks where his focus is on caching and multicast routing. He is also interested in optimization problems in general.
Bart Duysburgh received his electronic engineering degree in 1998 from the Ghent University, Belgium. Since September 1998, he is a research assistant for the Inter-University Micro-Electronics Center (IMEC) at the Department of Information Technology of the Ghent University. He worked on a software based IP and ATM traffic generator as part of his master thesis. His current research interests include active and programmable networks, traffic monitoring and network performance measurements, multicast routing, voice and video streaming over IP and assessment of voice and video quality.
Tim Wauters received his Masters Degree in Electrical Engineering in 2001 from the Ghent University, Belgium. He is now a Ph.D. student at the same University where he is part of the Department of Information Technology (INTEC). He works in the Integrated Broadband Communications Network group (IBCN). He worked on “Multicast Tree Optimalisation for Transcoding in Active Networks” as part of his graduate thesis. His current research interests include Peer-to-Peer Networks and Content Distribution Networks where his focus is on caching.
Filip De Turck received his M.Sc. degree in Electronic Engineering from the Ghent University, Belgium, in June 1997. In May 2002, he obtained the Ph.D. degree in Electronic Engineering from the same university. From October 1997 to September 2001, Filip De Turck was research assistant with the Fund for Scientific Research-Flanders, Belgium (F.W.O.-V.). At the moment, he is a part-time professor and a post-doctoral fellow of the F.W.O.-V., affiliated with the Department of Information Technology of the Ghent University. Filip De Turck is author or co-author of approximately 80 papers published in international journals or in the proceedings of international conferences. His main research interests include scalable software architectures for telecommunication network and service management, performance evaluation and optimization of routing, admission control and traffic management in telecommunication systems.
Bart Dhoedt received a degree in Engineering from the Ghent University in 1990. In September 1990, he joined the Department of Information Technology of the Faculty of Applied Sciences, University of Ghent. His research, addressing the use of micro-optics to realize parallel free space optical interconnects, resulted in a PhD degree in 1995. After a 2 year post-doc in opto-electronics, he became professor at the Faculty of Applied Sciences, Department of Information Technology. Since then, he is responsible for several courses on algorithms, programming and software development. His research interests are software engineering and mobile & wireless communications. Bart Dhoedt is author or co-author of approximately 100 papers published in international journals or in the proceedings of international conferences. His current research addresses software technologies for communication networks, peer-to-peer networks, mobile networks and active networks.
Piet Demeester received the Masters degree in Electro-technical engineering and the Ph.D degree from the Ghent University, Gent, Belgium in 1984 and 1988, respectively. In 1992 he started a new research activity on broadband communication networks resulting in the IBCN-group (INTEC Broadband communications network research group). Since 1993 he became professor at the Ghent University where he is responsible for the research and education on communication networks. The research activities cover various communication networks (IP, ATM, SDH, WDM, access, active, mobile), including network planning, network and service management, telecom software, internetworking, network protocols for QoS support, etc.
Piet Demeester is author of more than 400 publications in the area of network design, optimization and management. He is member of the editorial board of several international journals and has been member of several technical program committees (ECOC, OFC, DRCN, ICCCN, IZS, &).