Elsevier

Computer Networks

Volume 53, Issue 10, 14 July 2009, Pages 1587-1602
Computer Networks

An autonomic architecture for optimizing QoE in multimedia access networks

https://doi.org/10.1016/j.comnet.2008.11.004Get rights and content

Abstract

The recent emergence of multimedia services, such as Broadcast TV and Video on Demand over traditional twisted pair access networks, has complicated the network management in order to guarantee a decent Quality of Experience (QoE) for each user. The huge amount of services and the wide variety of service specifics require a QoE management on a per-user and per-service basis. This complexity can be tackled through the design of an autonomic QoE management architecture. In this article, the Knowledge Plane is presented as an autonomic layer that optimizes the QoE in multimedia access networks from the service originator to the user. It autonomously detects network problems, e.g. a congested link, bit errors on a link, etc. and determines an appropriate corrective action, e.g. switching to a lower bit rate video, adding an appropriate number of FEC packets, etc. The generic Knowledge Plane architecture is discussed, incorporating the triple design goal of an autonomic, generic and scalable architecture. The viability of an implementation using neural networks is investigated, by comparing it with a reasoner based on analytical equations. Performance results are presented of both reasoners in terms of both QoS and QoE metrics.

Introduction

In today’s broadband access networks, service operators are focusing on the deployment of new added value services such as Broadcast TV, Video on Demand (VoD), on-line gaming and Voice over IP (VoIP). Each of these services has large service demands: they often require a considerable amount of bandwidth and only tolerate a minimum amount of packet loss, delay and jitter. In order to meet these demands, current access networks are advancing from a best-effort service delivery to a QoS aware triple-play service delivery. Of prime importance in defining the quality of these services is the Quality of Experience (QoE): the quality as experienced by the end-user. The QoE degrades due to certain anomalies in their delivery (e.g. packet loss). The type of degradation depends highly on the type of service and the current context of the network. For example, an interactive service such as VoIP is typically more vulnerable to delay than a broadcast service. In the past, service operators have successfully deployed techniques such as VoD proxies to ensure a certain quality level between the service originator and the access node. However, the same techniques often fail when applied between the access node and the end-devices due to the heterogeneity of current home network configurations. Recently, these home networks are becoming complex networks consisting of different technologies (e.g. wired and wireless) and different user devices (e.g. set top boxes and traditional pc’s), with users concurrently accessing different services.

The focus of this article is on the development of an autonomic architecture to maximize the QoE of all services in multimedia access networks. As illustrated in Fig. 1, this architecture spans the complete network from service originator to the end-user. Its functionality is defined through three separate layers: the Monitor Plane (MPlane), the Knowledge Plane (KPlane) and the Action Plane (APlane). The MPlane monitors the network to gather information about the current situation, the KPlane uses this information to autonomously determine the best QoE optimizing action and the APlane applies the chosen actions into the network. A possible scenario for this architecture is the following: the MPlane observes that a video service is suffering from packet loss and informs the KPlane. Based on other monitor information, such as the amount of bandwidth used inside a network, the KPlane detects that each link in the network is still underutilized in terms of bandwidth and that only the services running over a specific link suffer from a drop in quality. The KPlane concludes that bit errors on that link must occur and instructs the APlane to add a specific amount of redundancy (through Forward Error Correction (FEC) packets) to tackle the packet loss measured.

In previous work, both the MPlane [1], [2] and APlane [3] architecture was studied. Also a first KPlane component was presented based on a set of analytical equations [3] and the use of fuzzy logic and neural networks to come to a more generic solution was explored [4]. Here, a generic and autonomic KPlane architecture to maximize the QoE of all services is proposed. Furthermore, a novel KPlane component is presented that is able to determine the best configuration of two QoE optimizing actions for a number of different scenarios. This novel component uses a neural network to make its decision. This component is compared to the earlier proposed analytical component. The analytical case is far from generic but performs very well and can be used as a reference to evaluate other solutions. The neural network approach is more generic and can have more advanced capabilities such as on-line learning behavior.

This article is structured as follows. First, in Section 2, some relevant work concerning autonomic computing and QoE optimization is discussed. In Section 4, the design objectives for the KPlane architecture are presented. Section 5 discusses the proposed architecture while Section 6 provides an overview of the two proposed solutions: the analytical reasoner (Section 6.2) and the neural network approach (Section 6.3). Next, in Section 7, results of both solutions obtained through simulation and a physical testbed are presented. Section 8, highlights how the architecture meets the design objectives and discusses the applicability of the two proposed solutions. Finally, Section 9 elaborates on future work.

Section snippets

Related work

The concept of a Knowledge Plane was originally presented in [5]. The authors argue that the fundamental design of the Internet, with its simple and transparent core with intelligence at the edges, leads to high management overhead. This is caused by the fact that the network simply forwards data packets, without knowing what its purpose is. While the edges can recognize that there is a problem with some service, the core cannot tell that something is wrong, because it has no idea of what

Relation between QoS and QoE of multimedia services

The QoE of multimedia services is mainly affected by the original quality of the multimedia service and the quality of its delivery. The first aspect is straightforward: audio and video services are encoded at a certain bitrate and these bitrate settings have a direct impact on the QoE of the services. A higher bitrate will result in a higher QoE and the service operator can decide at which QoE level he offers the service (e.g. a HD video as opposed to a SD video).

In our work, we focus on the

Design objectives

We have applied the Knowledge Plane [5] paradigm to optimize the QoE of multimedia services and extended it to a three plane approach consisting of a Knowledge Plane, a Monitor Plane and an Action Plane. The three planes form a fully autonomic loop that maps to the Monitor-Analyze-Plan-Execute (MAPE) control loop as originally presented by IBM [23]. The MPlane monitors service related information such as packet loss and router queue sizes through a set of monitor algorithms. By continuously

Architecture functional description

Fig. 2 illustrates the architecture of the three plane approach consisting of the earlier proposed Monitor Plane and Action Plane. In this article, the main focus is on the other two components in the architecture: a knowledge base comprising all relevant information and the Knowledge Plane that closes the autonomic loop through reasoning.

KPlane reasoning implementation

In this section, we will describe how we have implemented the reasoning behavior for a selected access network model. We have designed an analytical component that determines the correct actions to take by solving a number of mathematical equations. A second component makes use of neural networks to determine which actions to undertake.

Test setting

To test the performance of both reasoners we implemented the access network model as described in Section 6.1. The employed topology is illustrated in Fig. 5. In this topology two receivers can access different video services that are streamed from the video server over the access network. As explained in Section 6.1, the video server streams both a high quality (with an average bit rate of 2 Mbps) and a low quality (with an average bit rate of 500 kbps) video to the clients. We investigated

Discussion

As described in Section 5, an architecture was defined to optimize the QoE in multimedia access networks. The architecture forms a fully autonomic loop. No human interference is necessary to obtain a normal behavior as the architecture continuously monitors the network, reasons about which actions to take and executes these actions. Furthermore, as discussed in the previous section, a learning controller makes the architecture self-adaptive and introduces learning behavior. The third autonomic

Future work

Our work has mainly focused on a single Knowledge Plane instance on the access node, close to the user. This is a natural choice in a hierarchically, tree-like structured access network, where most problems can be solved locally. However, this limits the action range of the Knowledge Plane and can lead to suboptimal solutions. For example, when several clients are reporting packet loss for the same service, the cause is probably situated closer to the service originator and should be solved by

Conclusion

We defined an autonomic management architecture to optimize the QoE in multimedia access networks using a three plane approach. This three plane approach consists of a Monitor Plane that monitors the network and builds up knowledge about this network, a Knowledge Plane that analyzes the knowledge and determines the ideal QoE actions, and an Action Plane that enforces these actions into the network. We focused on the Knowledge Plane inside the architecture and defined how this architecture can

Acknowledgements

The research was performed partially within the framework of the MUSE-project, funded by the European Commission as part of the IST 6th Framework Program. The authors would like to thank all MUSE partners for their valuable contributions and feedback.

Steven Latré obtained a Masters degree in Computer Science from Ghent University, Belgium, in June 2006. Since August 2006, he is affiliated as a Ph.D. student with the Department of Information Technology at Ghent University. He is funded by grant of the Fund for Scientific Research of Flanders, Belgium (F.W.O.-V). His main research interest is the optimization of multimedia services through autonomic communications. He was involved in the IST FP6 project MUSE and is currently involved in the

References (38)

  • C. Wang et al.

    Application of neural networks for achieving 80211 QoS in heterogeneous channels

    Journal of Computer Networks

    (2008)
  • Z. Wang et al.

    Video quality assessment based on structural distortion measurement

    Signal Processing: Image Communication

    (2004)
  • T. Gruber

    A translation approach to portable ontology specifications

    Knowledge Acquisition

    (1993)
  • B. de Vleeschauwer, W. van de Meerssche, P. Simoens, F. de Turck, B. Dhoedt, P. Demeester, K. Struyve, T. van Caenegem,...
  • P. Simoens, B. de Vleeschauwer, W. van de Meerssche, F. de Turck, B. Dhoedt, P. Demeester, T. van Caeneghem, K....
  • B. de Vleeschauwer, P. Simoens, W. van de Meerssche, S. Latré, F. de Turck, P. Dhoedt, B. Demeester, S. van den Berghe,...
  • P. Simoens, S. Latré, B. de Vleeschauwer, W. van de Meerssche, F. de Turck, B. Dhoedt, P. Demeester, S. van den Berghe,...
  • D.D. Clark, C. Partridge, C.J. Ramming, J.T. Wroclawski, A Knowledge Plane for the Internet, in: SIGCOMM’03:...
  • P. Simoens, B. de Vleeschauwer, W. van de Meerssche, F. de Turck, B. Dhoedt, E. Gilon, K. Struyve, T. van Caenegem,...
  • D. Feldmeier et al.

    Protocol boosters

    IEEE Journal on Selected Areas in Communications

    (1998)
  • N. Feamster, H. Balakrishnan, Packet loss recovery for streaming video,...
  • M. Baglietto, G. Battistelli, R. Bolla, F.D.R. Bruschi, R. Zoppoli, A neural network solution to QoS-IP team-optimal...
  • R. del Hoyo-Alonso et al.

    Neural networks for QoS network management

    Lecture Notes in Computer Science, Computational and Ambient Intelligence

    (2007)
  • S. Mohamed et al.

    A study of real-time packet video quality using random neural networks

    IEEE Transactions on Circuits and Systems for Video Technology

    (2002)
  • M. Masugi, QoS mapping of VoIP communication using self-organizing neural networks, in: Proceedings of the IEEE...
  • D. de Vera, P. Rodrfguez-Bocca, G. Rubino, QoE monitoring platform for video delivery networks, in: Proceedings of the...
  • Y.J. Liang et al.

    Analysis of packet loss for compressed video: effect of burst losses and correlation between error frames

    IEEE Transactions on Circuits and Systems for Video Technology

    (2008)
  • A.R. Reibman, D. Poole, Characterizing packet-loss impairments in compressed video, in: Proceedings of the IEEE...
  • S.R. Gulliver, G. Ghinea, The perceptual influence of multimedia delay and jitter, in: Proceedings of the 2007 IEEE...
  • Cited by (0)

    Steven Latré obtained a Masters degree in Computer Science from Ghent University, Belgium, in June 2006. Since August 2006, he is affiliated as a Ph.D. student with the Department of Information Technology at Ghent University. He is funded by grant of the Fund for Scientific Research of Flanders, Belgium (F.W.O.-V). His main research interest is the optimization of multimedia services through autonomic communications. He was involved in the IST FP6 project MUSE and is currently involved in the EUREKA CELTIC project RUBENS.

    Pieter Simoens received his M.Sc. degree in Electronic Engineering from Ghent University, Belgium, in July 2005. Since October 2005, he is affiliated as research assistant with the Department of Information Technology of Ghent University. He is funded through the Fund for Scientific Research of Flanders, Belgium (F.W.O.-V.). His main research interest is the development of intelligent and adaptive network architectures and protocols designed for thin client computing and autonomous environments. He cooperated in the IST FP6 project MUSE and is now work package leader for the protocol design in the IST FP7 project MobiThin.

    Bart de Vleeschauwer obtained a Masters degree in Computer Science from the Ghent University, Belgium, in June 2003. Since August 2003, he is affiliated with the Department of Information Technology of the Ghent University. In 2008, he obtained the Ph.D. degree in Computer Science from the same university. He works there as a post-Ph.D. researcher. His research interests include the usage of overlay networks, Massively Multiplayer online Gaming and QoS monitoring. He was involved in the IST FP6 project MUSE and is currently involved in the EUREKA CELTIC project RUBENS.

    Wim Van de Meerssche received his M.Sc. degree in software development in 2004 from Ghent University, Belgium. In August 2004, he started working on software technologies for access networks in the Department of Information Technology (INTEC), at the same university. His work has been published in several scientific publications in international conferences. He was involved in the IST FP6 project MUSE and is currently involved in the EUREKA CELTIC project RUBENS.

    Filip De Turck received his M.Sc. degree in Electronic Engineering from Ghent University, Belgium, in June 1997. In May 2002, he obtained the Ph.D. degree in Electronic Engineering from the same university. At the moment, he is a full-time professor affiliated with the Department of Information Technology of Ghent University. His main research interests include scalable software architectures for telecommunication network and service management, performance evaluation and design of new telecommunication services.

    Bart Dhoedt received a degree in Engineering from Ghent University in 1990. In September 1990, he joined the Department of Information Technology, Ghent University. His research, addressing the use of micro-optics to realize parallel free space optical interconnects, resulted in a Ph.D. degree in 1995. After a 2-year post-doc in opto-electronics, he became Professor at the Department of Information Technology. He is responsible for courses on algorithms, programming and software development. His research interests are software engineering and mobile & wireless communications. 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 Ghent University, 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 Ghent University where he is responsible for the research and education on communication networks. The research activities cover various communication networks, including network planning, network and service management, telecom software, internetworking, network protocols for QoS support, etc.

    Steven van den Berghe graduated as Master in Computer Science at Ghent University in 1999. In the same year, he joined the Broadband Communication Network Group at Ghent University, where he performed research in the area of quality of service, traffic engineering and monitoring in IP networks, for which he was granted an IWT scholarship. This research resulted in the award of his Ph.D. in 2005. He joined the Alcatel-Lucent Bell Labs Wired Access team in 2006. His main topics of interest include design, control and management of multimedia and broadband service delivery networks.

    Edith Gilon de Lumley graduated in Electrical Engineering in 1998 at the UCL, Belgium. After her graduation she joined Alcatel-Lucent Bell and worked on Passive Optical Networks (PON) systems, especially the Physical Medium Dependent (PMD) layer from 1998, and she led the IST GIANT project from 2003 to 2005. She was coordinator of the IST MUSE sub-project B focusing on multimedia broadband.

    1

    Funded by a Ph.D. grant of the Fund for Scientific Research of Flanders (FWO-V).

    View full text