Research‎ > ‎

Management of Internet Applications

Today’s consumer Internet traffic is transmitted on a best effort basis without taking into account any quality requirements. The backbone and the wireless access networks lack service guarantees for the predominant consumer Internet traffic which is composed of applications like P2P or client-server file sharing, web browsing, or video streaming which together make up for more than 80 % of today’s traffic. In general the network does neither know which Internet applications it is carrying nor which quality requirements have to be met. To be able to meet the demands of applications and users in the network, QoE management requires an information exchange between application and network. From a conceptual perspective, QoE management requires three basic research steps: modeling QoE, monitoring QoE, optimizing QoE. This QoE Modeling, Monitoring, Optimization approach as introduced for cloud applications in general [11] and for the YouTube cloud in [7]. The optimization of video delivery via the Internet is discussed in Section , which also introduces QoE management for distributed video streaming based on scalable video codecs. Then, the optimization of overlay traffic resulting from Internet applications such as BitTorrent is considered in Section .

Video Streaming Services

The transmission of the video contents via the Internet is realized either with TCP or UDP. The usage of TCP guarantees the delivery of undisturbed video content since the protocol itself cares for the retransmissions of corrupted or lost packets. Further, it adapts the transport rate to network congestion, thus minimizing packet loss. If the available bandwidth is lower than the required video bit rate the video transmission lasts longer than the video playback. Thus, the playback is interrupted which is referred to as stalling. Hence, in case of TCP the video playback rather than the video itself is disturbed. In contrast, UDP does not perform bandwidth adaptation or guarantee packet delivery, but it transmits the data with the same bit rate as forwarded by the application. Thus, network congestion leads to lost packets which occur as artifacts or jumps in the stream. Hence, the user experiences a degraded video quality in terms of visual impairments. In order to compare the impact of the transport protocols on video streaming QoE in case of a bottleneck on the example of YouTube, two relevant steps are performed [9]. First, an intensive measurement study is conducted in order to quantify the relevant application-level QoS parameters over a bottleneck. In particular, the observed stalling patterns are modeled in terms of stalling frequency and stalling length. Second, YouTube video streaming via TCP and via UDP is compared from the end-user perspective by means of subjective user studies. The work is the first comparing QoE – and in particular YouTube QoE – for different transport protocols. It represents an important first step towards the appropriate selection of network protocols according to the demands and properties of Internet services based on the strict integration of the actual end user’s perspective. This QoE optimized selection may be realized e.g. by means of functional composition [8] or network virtualization [20, 19].

As a result of the QoE modeling process, QoE-relevant parameters are identified which have to bemonitored accordingly. For YouTube, [10] develops an appropriate QoE model for monitoring. The intention of the monitoring is to provide means for QoE management for ISPs or the video streaming service provider. From a QoE management perspective, stalling must be avoided to keep YouTube users satisfied. Even very short stalling events of a few seconds already decrease user perceived quality significantly. However, initial delays are tolerated up to a reasonable level. Hence, YouTube QoE monitoring should pro-actively detect imminent stalling, so that QoE control mechanisms are triggered timely in advance - which is possible with the provided QoE model in [10]. The passive monitoring approach in [18] detects YouTube video flows [7] and extracts video information from network packet data. Together with the YouTube video player parameters, derived in [9], the video buffer status is estimated almost exactly on behalf of network data only and QoE can be estimated. [2] devises and depicts evaluation results that show the potential and feasibility of doing real-time QoE monitoring in services such as YouTube in mobile broadband networks. The final step of QoE management aims at optimizing QoE in a controlled fashion. [7] optimizes YouTube QoE by avoiding stalling at costs of initial delay for prebuffering. This is necessary in the presence of insufficient network resources, e.g. for mobile YouTube [12]. An approximation for levels of initial delay is provided that are just high enough so that stalling is unlikely to occur. Possible implementation solution are discussed in [13, 3]for information exchange between network and application.

Additional management capabilities have to be introduced, however, if the video streaming users demand high-quality image resolutions that may require bandwidths greater than what is supported in the current Internet architecture. Massive investments by network and service providers are one pathway to cope with the emerging challenges. A QoE management approach is proposed in [24] which will lead to much more economic and efficient use of the available resource while improving QoE for end users. The QoE control mechanism takes into account the monitoring information and adjusts corresponding influence factors. For video streaming systems, the dynamic adaptation of the video quality according to the current situation can be smartly realized with the de facto state of the art video codec H.264 and its scalable extension (H.264/SVC). Guidelines how to control QoE depending on the actual network conditions are derived in [4, 25] quantifying the influence of video resolution, scaling method, video frame rate and video content types on the QoE by means of a measurement study in a laboratory environment using full-reference video metrics. The results found, e.g. that video sequences with lower resolution perform better than distorted high resolution content, are in line with the QoE provisioning-delivery hysteresis [4].

Content Distribution Overlays

The optimization of overlay traffic resulting from applications such as BitTorrent is a challenge addressed by several recent research initiatives. However, the assessment of such optimization techniques and their performance in the real Internet remains difficult. Despite a considerable set of works measuring real-life BitTorrent swarms, several characteristics of those swarms relevant for the optimization of overlay traffic have not yet been investigated. [6] overcomes this lack of realistic swarm statistics based on a large-scale distributed measurement campaign. The results in [6] show that real-life BitTorrent swarm distributions are highly skewed which is in particular true for regional swarms. The distribution of peers over ASs was found to follow a power-law. Further, the Pareto rule applies for the size of swarms, i.e. most of the peers (about 80 %) belong to the top 20 % of the swarms. Thus, a large optimization potential for locality awareness is revealed since these large swarms are (1) responsible for the majority of the BitTorrent traffic and (2) especially suitable for locality aware mechanisms. The concept of locality awareness is to optimize the traffic flow with information about the location of a content providing peer in the underlying network. In the current Internet, the application layer protocols are mostly unaware of the underlying network in accordance with the layered structure of the Internet’s protocol stack. Nevertheless, the need for improved network efficiency and the business interests of Internet service providers (ISPs) are both strong drivers towards a cross-layer approach in peer-to-peer protocol design, calling for P2P systems that would in some way interact with the ISPs. [3] gives an overview of the kinds of information that could potentially be exchanged between the P2P systems and the ISPs, and discusses their usefulness and the ease of obtaining and exchanging them. The possible approaches for interaction based on the level of involvement of the ISPs and the P2P systems are classified and evaluated in [3]. The influence of traffic management solutions on QoE for prevailing overlay applications is discussed in [21].

In [1] measurements of the distribution of a large number of live BitTorrent networks[6] are combined with the AS-level Internet topology. This allows to estimate in which tier of the Internet hierarchy BitTorrent traffic is mainly located and how much optimization potential exists for the different types of ISPs. Therewith, traffic flow and revenue implications of guiding Internet-wide BitTorrent swarms are analyzed. The results show that tier-1 ISPs profit from the un-managed exchange of P2P traffic and that these profits significantly decrease when the other ISPs would apply ALTO solutions. One solution approach is Economic Traffic Management (ETM) [5]. It operates under the assumption that all parties involved (ISP, user and, if applicable, overlay provider) will participate voluntarily in a management scheme they all profit from. As a consequence, this scheme has to provide incentives to these players to cooperate, so that in the end, a “TripleWin”, i.e., a win-win situation is created. Thus, ETM aims at reducing costs within the network while improving the QoE for end users. Two different ETM solution approaches are 1) locality promotion and 2) caching of contents via peers operated by ISPs. Locality promotion aims at reducing interdomain traffic by fostering the exchange of data among the peers within one domain. Therefore, topology information is exchanged between underlay and overlay resolving the information asymmetry. Studies from literature showed that existing locality approaches can lead to a win-win situation under certain conditions, and to a win-no lose situation in most cases. However, the scenarios used assume mostly homogeneous peer distributions and that all peers have the same access speed. This is not the case in practice according to the conducted BitTorrent measurement study[6]. Therefore, [16] extends previous work by studying scenarios with real-life, skewed peer distributions and heterogeneous access bandwidths of peers. The most important conclusions results are: (1) a win-no lose situation for ISPs and P2P users is difficult to achieve in practice with the current locality promotion proposals and (2) current proposals introduce or increase unfairness in the distribution process, in some cases they even decrease the overall efficiency of the distribution process. Thus, to summarize, current locality-aware peer selection mechanisms provide mainly a gain for the ISPs, but not for the end user. [17] proposes to group ASes which mitigates the effect of unbalanced download times. At the same time, it retains the full potential of inter-domain traffic savings for the ISPs. This is an important step towards a broad acceptance of locality-awareness in the P2P user community because P2P users can be sure that their AS affiliation has no impact on their application performance.

The second ETM approach aims at increasing QoE and reliability of a content distribution network by providing additional capacity supporting the dissemination of popular contents. It works by adding more resources to the overlay network in the form of ISP-controlled peers, so that the usage of these resources, such as storage and upload capacity, may be utilized to the ISPs advantage. [14, 15] considers how the different types of peer-to-peer caches, i.e. caches already available on the market and caches expected to become available in the future, can possibly affect the amount of inter-ISP traffic. As a result, caches can sometimes lead to increased outgoing transit traffic, depending on the portion of the peers within the ISP. Consequently, novel caching and neighbor selection policies were developed to countermeasure this effect. In general, ISP managed caches would be superior to transparent caches and to ISP managed ultrapeers in terms of decreasing the transit traffic.

For the future Internet, the area of socio-economics and incentives is being considered today as an important field of interest for network management. An indication of key questions and issues, which are classified as important for next steps in network management is derived in [22]. Furthermore, the process of dealing with aspects of “Socio-economic Management” is discussed, which determines a hybrid and innovative approach besides traditional network management approaches. This determines a network management in which control is delegated via socio-economic means and to a certain extend to the user and provider with the goal to maximize the overall social welfare and the networks technical efficiency at the same time. ETM approaches reflect a promising way for dealing with incentive issues being integrated into traditional network management approaches. This high level of attention for economic perspectives in network management is also witnessed in the book [23] constituting the thoroughly refereed proceedings of the 3rd International Workshop on Incentives, Overlays, and Economic Traffic Control.


[1] Valentin Burger, Frank Lehrieder, Tobias Hoßfeld, and Jan Seedorf. Who Profits from Peer-to-Peer File-Sharing? Traffic Optimization Potential in BitTorrent Swarms. In International Teletraffic Congress (ITC 24), Krakow, Poland, September 2012.

[2] Pedro Casas, Raimund Schatz, and Tobias Hoßfeld. Monitoring YouTube QoE: Is Your Mobile Network Delivering the Right Experience to your Customers? In IEEE Wireless Communications and Networking Conference (WCNC 2013), Shanghai, China, April 2013.

[3] György Dan, Tobias Hoßfeld, Simon Oechsner, Piotr Cholda, Rafal Stankiewicz, Ioanna Papafili, and George D. Stamoulis. Interaction Patterns between P2P Content Distribution Systems and ISPs. IEEE Communications Magazine, 49, May 2011.

[4] Tobias Hoßfeld, Markus Fiedler, and Thomas Zinner. The QoE Provisioning-Delivery-Hysteresis and Its Importance for Service Provisioning in the Future Internet . In Proceedings of the 7th Conference on Next Generation Internet Networks (NGI), Kaiserslautern, Germany, June 2011.

[5] Tobias Hoßfeld, David Hausheer, Fabio Hecht, Frank Lehrieder, Simon Oechsner, Ioanna Papafili, Peter Racz, Sergios Soursos, Dirk Staehle, George D. Stamoulis, Phuoc Tran-Gia, and Burkhard Stiller. An Economic Traffic Management Approach to Enable the TripleWin for Users, ISPs, and Overlay Providers. In Georgios Tselentis, John Domingue, Alex Galis, Anastasius Gavras, David Hausheer, Srdjan Krco, Volkmar Lotz, and Theodore Zahariadis, editors, FIA Prague Book, page 24. IOS Press Books Online, Towards the Future Internet - A European Research Perspective, May 2009.

[6] Tobias Hoßfeld, Frank Lehrieder, David Hock, Simon Oechsner, Zoran Despoto-vic, Wolfgang Kellerer, and Maximilian Michel. Characterization of BitTorrent Swarms and their Distribution in the Internet. Computer Networks, 55, April 2011.

[7] Tobias Hoßfeld, Florian Liers, Raimund Schatz, Barbara Staehle, Dirk Staehle, Thomas Volkert, and Florian Wamser. Quality of Experience Management for YouTube: Clouds, FoG and the AquareYoum. PIK - Praxis der Informationverarbeitung und -kommunikation (PIK), 35, August 2012.

[8] Tobias Hoßfeld, Florian Liers, Thomas Volkert, and Raimund Schatz. FoG and Clouds: Optimizing QoE for YouTube. In KuVS 5thGI/ITG KuVS Fachgespräch NG Service Delivery Platforms, Munich, Germany, October 2011.

[9] Tobias Hoßfeld and Raimund Schatz. Quality of experience of youtube video streaming for current internet transport protocols. Computer Networks, 2012.

[10] Tobias Hoßfeld, Raimund Schatz, Ernst Biersack, and Louis Plissonneau. Internet Video Delivery in YouTube: From Traffic Measurements to Quality of Experience. In Ernst Biersack, Christian Callegari, and Maja Matijasevic, editors, Data Traffic Monitoring and Analysis: From measurement, classification and anomaly detection to Quality of experience, Lecture Notes in Computer Science - Computer Communication Networks and Telecommunications. Springer, 2013 in press.

[11] Tobias Hoßfeld, Raimund Schatz, Martin Varela, and Christian Timmerer. Challenges of QoE Management for Cloud Applications. IEEE Communications Magazine, 50, April 2012.

[12] Tobias Hoßfeld, Dominik Strohmeier, Alexander Raake, and Raimund Schatz. Pippi Longstocking Calculus for Temporal Stimuli Pattern on YouTube QoE. In 5th ACM Workshop on Mobile Video (MoVid 2013), Oslo, Norway, February 2013.

[13] Tobias Hoßfeld, Thomas Zinner, Markus Fiedler, Florian Liers, Thomas Volkert, Rahamatullah Khondoker, and Raimund Schatz. Requirement Driven Prospects for Realizing User-Centric Network Orchestration. Computer Communications Journal Special Issue on Human-Centric Multimedia Networking, 2012.

[14] Frank Lehrieder, György Dan, Tobias Hoßfeld, Simon Oechsner, and Vlad Singeorzan. The Impact of Caching on BitTorrent-like Peer-to-peer Systems. In 10th IEEE International Conference on Peer-to-Peer Computing 2010 - IEEE P2P 2010, Best Paper Award, Delft, the Netherlands, August 2010.

[15] Frank Lehrieder, György Dan, Tobias Hoßfeld, Simon Oechsner, and Vlad Singeorzan. Caching for BitTorrent-like P2P Systems: A Simple Fluid Model and its Implications. IEEE/ACM Transactions on Networking (TON), 20, August 2012.

[16] Frank Lehrieder, Simon Oechsner, Tobias Hoßfeld, Zoran Despotovic, Wolfgang Kellerer, and Maximilian Michel. Can P2P-Users Benefit from Locality-Awareness? In10th IEEE International Conference on Peer-to-Peer Computing 2010 - IEEE P2P 2010, Delft, the Netherlands, August 2010.

[17] Frank Lehrieder, Simon Oechsner, Tobias Hoßfeld, Dirk Staehle, Zoran Despotovic, Wolfgang Kellerer, and Maximilian Michel. Mitigating Unfairness in Locality-Aware Peer-to-Peer Networks. International Journal of Network Management (IJNM), Special Issue on Economic Traffic Management, 21, January 2011.

[18] Raimund Schatz, Tobias Hoßfeld, and Pedro Casas. Passive YouTube QoE Monitoring for ISPs. In Best Paper Award at Workshop on Future Internet and Next Generation Networks (FINGNet-2012), Palermo, Italy, July 2012.

[19] D. Schlosser, M. Hoffmann, T. Hoßfeld, M. Jarschel, A. Kirstaedter, W. Kellerer, and S. Köhler. COMCON: Use Cases for Virtual Future Networks. In TridentCom 2010, Berlin, May 2010.

[20] Daniel Schlosser, Michael Jarschel, Michael Duelli, Tobias Hoßfeld, Klaus Hoffmann, Marco Hoffmann, Hans Jochen Morper, Dan Jurca, and Ashiq Khan. A Use Case Driven Approach to Network Virtualization. In accepted at IEEE Kaleidoscope 2010, published via OPUS Würzburg under OpenAccess, Würzburg, Germany, December 2010.

[21] Rafal Stankiewicz, Piotr Cholda, Jerzy Domzal, Robert Wojcik, Tobias Hoßfeld, Thomas Zinner, Simon Oechsner, and Frank Lehrieder. Influence of traffic management solutions on Quality of Experience for prevailing overlay applications . InProceedings of the 7th Conference on Next Generation Internet Networks (NGI), Kaiserslautern, Germany, June 2011.

[22] Burkhard Stiller, David Hausheer, Tobias Hoßfeld, Antonio Liotta, Spiros Spirou, and Martin Waldburger. A Discussion of Socio-economic Management and Incentives for the Future Internet. In 2nd International Workshop on the Network of the Future (FutureNet II) in conjunction with IEEE GLOBECOM 2009, Honolulu, Hawaii, December 2009.

[23] Burkhard Stiller, Tobias Hoßfeld, and George D. Stamoulis. Incentives, Overlays, and Economic Traffic Control. Lecture Notes in Computer Science - Computer Communication Networks and Telecommunications. Springer, Proceedings of the Third Workshop on Economic Traffic Management (ETM 2010), Amsterdam, The Netherlands, August 2010.

[24] Thomas Zinner, Osama Abboud, Oliver Hohlfeld, Tobias Hoßfeld, and Phuoc Tran-Gia. Towards QoE Management for Scalable Video Streaming. In 21th ITC Specialist Seminar on Multimedia Applications - Traffic, Performance and QoE, Miyazaki, Jap, March 2010.

[25] Thomas Zinner, Oliver Hohlfeld, Osama Abboud, and Tobias Hoßfeld. Impact of Frame Rate and Resolution on Objective QoE Metrics. In International Workshop on Quality of Multimedia Experience, 2010, Trondheim, June 2010.