ABSTRACT
Major commercial client-side video players employ adaptive bitrate (ABR) algorithms to improve user quality of experience (QoE). With the evolvement of ABR algorithms, increasingly complex methods such as neural networks have been adopted to pursue better performance. However, these complex methods are too heavyweight to be directly implemented in client devices, especially mobile phones with very limited resources. Existing solutions suffer from a trade-off between algorithm performance and deployment overhead. To make the implementation of sophisticated ABR algorithms practical, we propose PiTree, a general, high-performance and scalable framework that can faithfully convert sophisticated ABR algorithms into lightweight decision trees to reduce deployment overhead. We also provide a theoretical upper bound on the optimization loss during the conversion. Evaluation results on three representative ABR algorithms demonstrate that PiTree could faithfully convert ABR algorithms into decision trees with <3% average performance degradation. Moreover, comparing to original implementation solutions, PiTree could save operating expenses for large content providers.
- Chrome devtools. https://developers.google.com/web/tools/chrome-devtools/.Google Scholar
- Cplex optimizer. https://www.ibm.com/analytics/cplex-optimizer.Google Scholar
- Ec2 instance pricing -- amazon web services (aws). https://aws.amazon.com/ec2/pricing/reserved-instances/pricing/.Google Scholar
- Mobile accounted for 62 percent of online video views. https://www.statista.com/statistics/444318/mobile-device-video-views-share/.Google Scholar
- Official youtube blog: With nearly 2 million concurrent viewers and over 3 million live watch hours, first presidential debate breaks political record. https://youtube.googleblog.com/2016/09/with-nearly-2-million-concurrent.html.Google Scholar
- Tornado web server. https://www.tornadoweb.org/.Google Scholar
- tsenart/vegeta: Http load testing tool and library. it's over 9000! https://github.com/tsenart/vegeta.Google Scholar
- Youtube. https://www.youtube.com/.Google Scholar
- Youtube revenue and usage statistics (2018) - business of apps. http://www.businessofapps.com/data/youtube-statistics/.Google Scholar
- Raw data - measuring broadband america. https://www.fcc.gov/reports-research/reports/measuring-broadband-america/raw-data-measuring-broadband-america-2016, 2016.Google Scholar
- Dash.js. https://github.com/Dash-Industry-Forum/dash.js, 2018.Google Scholar
- gruntjs/grunt-contrib-uglify: Minify files with uglifyjs. https://github.com/gruntjs/grunt-contrib-uglify, 2019.Google Scholar
- Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., Devin, M., Ghemawat, S., et al. Tensorflow: A system for large-scale machine learning. In USENIX OSDI (2016).Google ScholarDigital Library
- Akhtar, Z., Nam, Y. S., Govindan, R., Rao, S., Chen, J., Katz-Bassett, E., Ribeiro, B., Zhan, J., and Zhang, H. Oboe: auto-tuning video abr algorithms to network conditions. In ACM SIGCOMM (2018).Google ScholarDigital Library
- Ba, J., and Caruana, R. Do deep nets really need to be deep? In NIPS (2014).Google Scholar
- Bastani, O., Pu, Y., and Solar-Lezama, A. Verifiable reinforcement learning via policy extraction. In NeurIPS (2018).Google Scholar
- Beben, A., Wisniewski, P., Batalla, J. M., and Krawiec, P. AbmaGoogle Scholar
- : lightweight and efficient algorithm for http adaptive streaming. In ACM MMSys (2016).Google Scholar
- Ben Mustafa, I., Nadeem, T., and Halepovic, E. Flexstream: Towards flexible adaptive video streaming on end devices using extreme sdn. In ACM MM (2018), pp. 555--563.Google ScholarDigital Library
- Bentaleb, A., Begen, A. C., Harous, S., and Zimmermann, R. A distributed approach for bitrate selection in http adaptive streaming. In ACM MM (2018), pp. 573--581.Google ScholarDigital Library
- Blockeel, H., and De Raedt, L. Top-down induction of first-order logical decision trees. Artificial intelligence 101, 1--2 (1998), 285--297.Google Scholar
- Bucilua, C., Caruana, R., and Niculescu-Mizil, A. Model compression. In ACM KDD (2006).Google ScholarDigital Library
- Chen, W., Wilson, J., Tyree, S., Weinberger, K., and Chen, Y. Compressing neural networks with the hashing trick. In ICML (2015).Google ScholarDigital Library
- Chen, Y.-H., Krishna, T., Emer, J. S., and Sze, V. Eyeriss: An energy-efficient reconfigurable accelerator for deep convolutional neural networks. IEEE Journal of Solid-State Circuits 52, 1 (2017), 127--138.Google ScholarCross Ref
- Cui, Y., Ooi, W. T., Liu, J., Zhang, X., Bentaleb, A., Zheng, K., and Li, Y. Acm multimedia 2019 grand challenge - live video streaming. https://www.aitrans.online/MMGC/, 2019.Google Scholar
- De Cicco, L., Caldaralo, V., Palmisano, V., and Mascolo, S. Elastic: A client-side controller for dynamic adaptive streaming over http (dash). In IEEE International Packet Video Workshop (2013).Google ScholarCross Ref
- De Cicco, L., Cilli, G., and Mascolo, S. Erudite: a deep neural network for optimal tuning of adaptive video streaming controllers. In ACM MMSys (2019).Google ScholarDigital Library
- Dobrian, F., Sekar, V., Awan, A., Stoica, I., Joseph, D., Ganjam, A., Zhan, J., and Zhang, H. Understanding the impact of video quality on user engagement. In ACM SIGCOMM (2011).Google ScholarDigital Library
- Friedman, J. H., Olshen, R. A., Stone, C. J., et al. Classification and regression trees. Belmont, CA: Wadsworth & Brooks (1984).Google Scholar
- Gama, J., Rocha, R., and Medas, P. Accurate decision trees for mining high-speed data streams. In ACM KDD (2003).Google ScholarDigital Library
- Ganjam, A., Siddiqui, F., Zhan, J., Liu, X., Stoica, I., Jiang, J., Sekar, V., and Zhang, H. C3: Internet-scale control plane for video quality optimization. In USENIX NSDI (2015), pp. 131--144.Google ScholarDigital Library
- Guo, W., Mu, D., Xu, J., Su, P., Wang, G., and Xing, X. Lemna: Explaining deep learning based security applications. In ACM CCS (2018).Google ScholarDigital Library
- Huang, T., Yao, X., Wu, C., Zhang, R.-X., and Sun, L. Tiyuntsong: A self-play reinforcement learning approach for abr video streaming. arXiv:1811.06166 (2018).Google Scholar
- Huang, T.-Y., Handigol, N., Heller, B., McKeown, N., and Johari, R. Confused, timid, and unstable: Picking a video streaming rate is hard. In ACM IMC (2012), pp. 225--238.Google ScholarDigital Library
- Huang, T.-Y., Johari, R., McKeown, N., Trunnell, M., and Watson, M. A buffer-based approach to rate adaptation: Evidence from a large video streaming service. In ACM SIGCOMM (2014).Google ScholarDigital Library
- Hussein, A., Gaber, M. M., Elyan, E., and Jayne, C. Imitation learning: A survey of learning methods. ACM Comput. Surv. 50, 2 (Apr. 2017), 21:1--21:35.Google ScholarDigital Library
- Index, C. V. N. Forecast and methodology, 2016--2021. White Paper, June (2017).Google Scholar
- Jiang, J., Sekar, V., Milner, H., Shepherd, D., Stoica, I., and Zhang, H. Cfa: A practical prediction system for video qoe optimization. In USENIX NSDI (2016), pp. 137--150.Google Scholar
- Jiang, J., Sekar, V., and Zhang, H. Improving fairness, efficiency, and stability in http-based adaptive video streaming with festive. In ACM CoNEXT (2012), pp. 97--108.Google ScholarDigital Library
- Kakade, S. M., and Tewari, A. On the generalization ability of online strongly convex programming algorithms. In NIPS (2008).Google Scholar
- Koller, R., and Williams, D. Will serverless end the dominance of linux in the cloud? In ACM HotOS (2017), pp. 169--173.Google ScholarDigital Library
- Krishnan, S. S., and Sitaraman, R. K. Video stream quality impacts viewer behavior: Inferring causality using quasi-experimental designs. In ACM IMC (2012), pp. 211--224.Google ScholarDigital Library
- Lane, N. D., Georgiev, P., and Qendro, L. Deepear: Robust smartphone audio sensing in unconstrained acoustic environments using deep learning. In ACM Ubicomp (2015).Google ScholarDigital Library
- Li, H., Kadav, A., Durdanovic, I., Samet, H., and Graf, H. P. Pruning filters for efficient convnets. In ICLR (2017).Google Scholar
- Li, Z., Zhu, X., Gahm, J., Pan, R., Hu, H., Begen, A. C., and Oran, D. Probe and adapt: Rate adaptation for http video streaming at scale. IEEE Journal on Selected Areas in Communications 32, 4 (2014), 719--733.Google ScholarCross Ref
- Liu, J., Tao, X., and Lu, J. Qoe-oriented rate adaptation for dash with enhanced deep q-learning. IEEE Access 7 (2019), 8454--8469.Google ScholarCross Ref
- Lundberg, S. M., and Lee, S.-I. A unified approach to interpreting model predictions. In NIPS (2017).Google Scholar
- Mao, H., Netravali, R., and Alizadeh, M. Neural adaptive video streaming with pensieve. In ACM SIGCOMM (2017), pp. 197--210.Google ScholarDigital Library
- Netravali, R., Sivaraman, A., Das, S., Goyal, A., Winstein, K., Mickens, J., and Balakrishnan, H. Mahimahi: Accurate record-and-replay for HTTP. In USENIX ATC (2015), pp. 417--429.Google Scholar
- Ramos-Mu noz, J. J., Prados-Garzon, J., Ameigeiras, P., Navarro-Ortiz, J., and López-Soler, J. M. Characteristics of mobile youtube traffic. IEEE Wireless Communications 21, 1 (2014), 18--25.Google Scholar
- Ribeiro, M. T., Singh, S., and Guestrin, C. Why should i trust you?: Explaining the predictions of any classifier. In ACM KDD (2016).Google ScholarDigital Library
- Riiser, H., Vigmostad, P., Griwodz, C., and Halvorsen, P. Commute path bandwidth traces from 3g networks: Analysis and applications. In ACM MMSys (2013), pp. 114--118.Google ScholarDigital Library
- Ross, S., Gordon, G., and Bagnell, D. A reduction of imitation learning and structured prediction to no-regret online learning. In AISTATS (2011).Google Scholar
- Sengupta, S., Ganguly, N., Chakraborty, S., and De, P. Hotdash: Hotspot aware adaptive video streaming using deep reinforcement learning. In IEEE ICNP (2018), pp. 165--175.Google ScholarCross Ref
- Smilkov, D., Thorat, N., Assogba, Y., Nicholson, C., Kreeger, N., Yu, P., Cai, S., Nielsen, E., et al. Tensorflow.js: Machine learning for the web and beyond (https://js.tensorflow.org/). In SysML (2019).Google Scholar
- Spiteri, K., Sitaraman, R., and Sparacio, D. From theory to practice: improving bitrate adaptation in the dash reference player. In ACM MMSys (2018).Google ScholarDigital Library
- Spiteri, K., Urgaonkar, R., and Sitaraman, R. K. Bola: Near-optimal bitrate adaptation for online videos. In IEEE INFOCOM (2016).Google Scholar
- Stockhammer, T. Dynamic adaptive streaming over http -- standards and design principles. In ACM MMSys (2011), pp. 133--144.Google Scholar
- Sun, Y., Yin, X., Jiang, J., Sekar, V., Lin, F., Wang, N., Liu, T., and Sinopoli, B. Cs2p: Improving video bitrate selection and adaptation with data-driven throughput prediction. In ACM SIGCOMM (2016).Google ScholarDigital Library
- Verma, A., Murali, V., Singh, R., Kohli, P., and Chaudhuri, S. Programmatically interpretable reinforcement learning. In ICML (2018).Google Scholar
- Wang, B., and Ren, F. Towards forward-looking online bitrate adaptation for dash. In ACM MM (2017), pp. 1122--1129.Google ScholarDigital Library
- Wang, C., Rizk, A., and Zink, M. Squad: A spectrum-based quality adaptation for dynamic adaptive streaming over http. In ACM MMSys (2016).Google ScholarDigital Library
- Wu, J., Cheng, B., Yuen, C., Shang, Y., and Chen, J. Distortion-aware concurrent multipath transfer for mobile video streaming in heterogeneous wireless networks. IEEE Transactions on Mobile Computing 14, 4 (2015), 688--701.Google ScholarDigital Library
- Yadav, P. K., Shafiei, A., and Ooi, W. T. Quetra: A queuing theory approach to dash rate adaptation. In ACM MM (2017).Google ScholarDigital Library
- Yeo, H., Jung, Y., Kim, J., Shin, J., and Han, D. Neural adaptive content-aware internet video delivery. In USENIX OSDI (2018).Google Scholar
- Yin, X., Jindal, A., Sekar, V., and Sinopoli, B. A control-theoretic approach for dynamic adaptive video streaming over http. In ACM SIGCOMM (2015).Google ScholarDigital Library
- Zheng, Y., Liu, Z., You, X., Xu, Y., and Jiang, J. Demystifying deep learning in networking. In ACM APNet (2018).Google ScholarDigital Library
Index Terms
- PiTree: Practical Implementation of ABR Algorithms Using Decision Trees
Recommendations
ABR Service in ATM Networks: Performance Comparison between Adaptive Stochastic, ERICA, BECN and BP Congestion Control Schemes with Guaranteed Minimum Cell Rate
IC3N '98: Proceedings of the International Conference on Computer Communications and NetworksPerformances of four congestion control schemes for ABR service in ATM networks: Explicit Rate indication for Congestion Avoidance (ERICA), Backward Explicit Congestion Notification (BECN), Back Pressure (BP) and proposed Adaptive Stochastic (AS) are ...
SERICA _ω. an extension of the SERICA algorithm
This work extends a recently developed scheme, namely the Simple Explicit Rate Identification with Congestion Avoidance (SERICA) algorithm, which regulates the ABR traffic flow in an ATM network. The new scheme (SERICA_ω), like SERICA, takes into ...
Modeling ATM systems with GSPNs and SWNs
Special issue on Stochastic Petri NetsThis paper overviews the work of the authors in the field of modeling and analysis of Asynchronous Transfer Mode (ATM) networks using Generalized Stochastic Petri Nets (GSPN) and a special class of high-level stochastic Petri nets known as Stochastic ...
Comments