Distributed DASH Dataset
Download as PPTX, PDF2 likes14,852 views
The document describes a distributed Dynamic Adaptive Streaming over HTTP (DASH) dataset that enables objective comparison of DASH client implementations. The dataset contains a 1 hour and 37 minute video segmented into multiple representations and distributed across multiple servers. Researchers and developers are invited to mirror the dataset and register their server to participate in the distributed dataset. The distributed nature allows evaluation of clients that switch between servers. The dataset can be used to study issues like startup delay and bandwidth fluctuations across servers.
![Main repository and distribution
• Available at http://dash.itec.aau.at | http://bit.ly/d-dash
• RedBull Playstreet sequence, 1h 37min 28sec
– Segment length: 2, 4, 6, 10, 15sec
– 17 different video representations: [100kbps at 320x240, 6
Mbps at 1920x1080]
– 4 different audio representations: two channels at
64, 96, 128, and 165 kbps using a 48 kHz sampling rate
Feb 28, 2013 ACM MMSys 2013 6](https://image.slidesharecdn.com/mmsysddash-130228165220-phpapp01/85/Distributed-DASH-Dataset-6-320.jpg)
Distributed DASH Dataset
- 1. Distributed DASH Dataset Stefan Lederer, Christopher Müller, Christian Timmerer, Cyril Concolato, Jean Le Feuvre, and Karel Fliegel Alpen-Adria-Universität Klagenfurt (AAU) Faculty of Technical Sciences (TEWI) Department of Information Technology (ITEC) Multimedia Communication (MMC) Sensory Experience Lab (SELab) http://research.timmerer.com http://blog.timmerer.com http://dash.itec.aau.at/mailto:christian.timmerer@itec.uni-klu.ac.at ACM Multimedia Systems 28th February, 2013 Acknowledgments. This work was supported in part by the EC in the context of the ALICANTE (FP7-ICT-248652) and SocialSensor (FP7-ICT- 287975) projects and partly performed in the Lakeside Labs research cluster at AAU. Special thanks to the Red Bull Media House for providing us the Red Bull Playstreets video. They own the rights of the content but the usage for scientific purposes is permitted. This work was also supported in part by the French-funded project AUSTRAL (DGCIS FUI13). This work was partially supported by the COST IC1003 QUALINET, by the Czech-funded project COST CZ LD12018 MOVERIQ and by the grant of the Czech Science Foundation No. P102/10/1320.
- 2. What is DASH? http://en.wikipedia.org/wiki/Dash_(disambiguation) Feb 28, 2013 ACM MMSys 2013 2
- 3. Dynamic Adaptive Streaming over HTTP • In a nutshell … Adaptation logic is within the client, not normatively specified by the standard, subject to research and development Christian Timmerer and Carsten Griwodz. 2012. Dynamic adaptive streaming over HTTP: from content creation to consumption. In Proceedings of the 20th ACM international conference on Multimedia (MM '12). ACM, New York, NY, USA, 1533-1534. DOI=10.1145/2393347.2396553 http://doi.acm.org/10.1145/2393347.2396553 http://www.slideshare.net/christian.timmerer/dynamic-adaptive-streaming-over-http-from-content-creation-to-consumption Feb 28, 2013 ACM MMSys 2013 3
- 4. Why a (distributed) DASH dataset? • … to enable an objective comparison of evaluation results across different client implementations – E.g.: MMSys dataset track, QUALINET database Lederer, S., Mueller, C., and Timmerer, C. 2012. Dynamic adaptive streaming over HTTP dataset. In Proceedings of the 3rd Multimedia Systems Conference (MMSys '12). ACM, New York, NY, USA, 89-94. DOI=http://doi.acm.org/10.1145/2155555.2155570 • Why distributed? – DASH allows to pull segments from multiple sources/sites – Signaled through multiple BaseURL elements within the XML-based Media Presentation Description (MPD) – Allows for a real-world evaluation of DASH clients that perform bitstream switching between multiple sites – E.g., to simulate switching between multiple Content Distribution Networks (CDNs) • Additionally, we provide a mechanism to mirror the DASH content to further sites – Join this activity, everyone is invited – get involved in and exited about DASH! Feb 28, 2013 ACM MMSys 2013 4
- 5. DASH and multiple BaseURLs • BaseURL – URL indicating a location that can be used to request the different segments needed for the presentation – Optional element, can be present multiple times at multiple levels in the XML hierarchy of the MPD – Optional attributes: serviceLocation and byteRange Feb 28, 2013 ACM MMSys 2013 5
- 6. Main repository and distribution • Available at http://dash.itec.aau.at | http://bit.ly/d-dash • RedBull Playstreet sequence, 1h 37min 28sec – Segment length: 2, 4, 6, 10, 15sec – 17 different video representations: [100kbps at 320x240, 6 Mbps at 1920x1080] – 4 different audio representations: two channels at 64, 96, 128, and 165 kbps using a 48 kHz sampling rate Feb 28, 2013 ACM MMSys 2013 6
- 7. Add your site to the D-DASH dataset • Create a mirror of the dataset: – Copy the dataset to your server and provide HTTP-access to it. The dataset has a size of approx. 85 GB and can be downloaded from our servers: FTP: ftp://ftp-itec.uni-klu.ac.at/pub/datasets/mmsys13/ HTTP: http://www-itec.uni-klu.ac.at/ftp/datasets/mmsys13/ – It is recommended to create a job, e.g., via wget, to keep the mirror up-to-date and the get latest MPDs also on your site. This can be done via the the following command line: wget -m -nH –cut-dirs=3 ftp://ftp-itec.uni-klu.ac.at/pub/datasets/mmsys13/ • Register the mirror of the dataset: – Please register your site so that we can validate your dataset copy and add your site to the MPDs of the dataset – Please use our registration form at: http://www-itec.uni-klu.ac.at/dash/ddash/register.html • You are part of D-DASH! – After the registration we check your dataset mirror and you will be notified by us. Your site will be added to the MPDs in our dataset repository and mirrored to all other sites. – Furthermore your site will be integrated in our MPD-generation service Feb 28, 2013 ACM MMSys 2013 7
- 8. MPD update process Method 1 • MPDs of the dataset are updated in the main repository – If new mirrors have been added and verified – If an existing mirror gets inactive • These MPDs are replicated but its the responsibility of the site owners Method 2 • MPD generation service (PHP script) which provides the most up-to-date MPDs based on our mirror database – http://www-itec.uni- klu.ac.at/dash/ddash/mpdGenerator.php?segmentlength={2, 4, 6, 10, 15}&type={full, URLTemplate} Feb 28, 2013 ACM MMSys 2013 8
- 9. What can you do with this dataset? • Work on a paper! E.g., for QoMEX’13 (submission deadline Mar. 6/20), JSAC Special Issue (Apr. 1), PV’13 (June), or MMSys’14 (Sep. 16) • Bootstrap problem – When retrieving an MPD with multiple BaseURLs, with which BaseURL to start a DASH session? – Finding the “best” BaseURL to use may influence the start-up delay and, thus, Quality of Experience • Bandwidth fluctuations during a DASH session – Switch to another BaseURL (which one?) or select another representation within the same BaseURL • Live streaming with multiple BaseURL – well, that’s another story! • You may use “Commute Path Bandwidth Traces from 3G Networks: Analysis and Applications” from Riiser et al. Feb 28, 2013 ACM MMSys 2013 9
- 10. Conclusions • Major critical issue for DASH implementations – Bandwidth estimations for segments @ multiple BaseURLs in parallel – Subject to low start-up delay and smooth streaming without stalls or re-buffering • Our distributed DASH dataset allows for a real-world evaluation of DASH clients that perform bitstream switching between multiple sites • Current sites: Klagenfurt (Austria), Paris (France), Prague (Czech Republic) • It can be easily distributed further, e.g., outside Europe Feb 28, 2013 ACM MMSys 2013 10
- 11. http://multimediacommunication.blogspot.no/2012/07/jsac-special-issue-adaptive-media.html Guest Editors • Christian Timmerer, Alpen-Adria-Universität Klagenfurt, Austria • Ali C. Begen, CISCO, Canada • Thomas Stockhammer, QUALCOMM, USA • Carsten Griwodz, Simula Research Laboratory, Norway • Bernd Girod, Stanford University, USA Feb 28, 2013 ACM MMSys 2013 11
- 12. Thank you for your attention ... questions, comments, etc. are welcome … Ass.-Prof. Dipl.-Ing. Dr. Christian Timmerer Klagenfurt University, Department of Information Technology (ITEC) Universitätsstrasse 65-67, A-9020 Klagenfurt, AUSTRIA christian.timmerer@itec.uni-klu.ac.at http://research.timmerer.com/ Tel: +43/463/2700 3621 Fax: +43/463/2700 3699 © Copyright: Christian Timmerer Feb 28, 2013 ACM MMSys 2013 12