CS2P: Improving Video Bitrate Selection and Adaptation with Data-Driven Throughput Prediction
Open accessThe DOI matches its Crossref record; 38 of the 38 references it cites were found in a registry or an index.
Abstract
Bitrate adaptation is critical in ensuring good users’ quality-of-experience (QoE) in Internet video delivery system. Several efforts have argued that accurate throughput prediction can dramatically improve (1) initial bitrate selection for low startup delay and high initial resolution; (2) midstream bitrate adaptation for high QoE. However, prior ef- forts did not systematically quantify real-world throughput predictability or develop good prediction algorithms. To bridge this gap, this paper makes three key technical contributions: First, we analyze the throughput characteristics in a dataset with 20M+ sessions. We find: (a) Sessions sharing similar key features (e.g., ISP, region) present similar initial values and dynamical patterns; (b) There is a natural “stateful” dynamical behavior within a given session. Second, building on these insights, we develop CS2P, a better throughput prediction system. CS2P leverages data-driven approach to learn (a) clusters of similar sessions, (b) an initial throughput predictor, and (c) a Hidden-Markov-Model based midstream predictor modeling the stateful evolution of throughput. Third, we develop a prototype system and show by trace-driven simulation and real-world experiments that CS2P outperforms state-of-art by 40% and 50% median pre- diction error respectively for initial and midstream through- put and improves QoE by 14% over buffer-based adaptation algorithm.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“And deep learning further empowers more effective local analysis and prediction [130] at the edge node of industry instead of the cloud.”
From Deep learning for edge computing applications: A State-of-the-Art survey (2020) · cites this paper as [130]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“It has been shown in recent work [432] that accurate throughput prediction can significantly improve the QoE for adaptive video streaming.”
From A comprehensive survey on machine learning for networking: Evolution, applications and research opportunities (2018) · cites this paper as [432]
The cited source says what the claim attributes to it.
In the citing paper
“It is also shown that CS2P achieves 3.2% improvement on overall QoE and 10.9% higher average bitrate over state-of-art Model Predictive Control (MPC) approach, which uses harmonic mean for throughput prediction.”
From A comprehensive survey on machine learning for networking: Evolution, applications and research opportunities (2018) · cites this paper as [432]
In the citing paper
“Data-driven approaches, in particular ML, have been employed mainly in two different ways: (1) to predict changes in network QoS, namely throughput, and trigger adaptation mechanism to reduce rebuffering time [432], and (2) to select appropriate adaptation action [102].”
From A comprehensive survey on machine learning for networking: Evolution, applications and research opportunities (2018) · cites this paper as [432]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Book Review: Pattern Recognition and Machine Learning
Nasrabadi, N. M. (2007) · Journal of Electronic Imaging · doi:10.1117/1.2819119 · Cited by 22,066
Pattern Recognition and Machine Learning
(2006) · doi:10.1007/978-0-387-45528-0 · Cited by 9,824
Scikit-learn: Machine Learning in Python
(2011) · Journal of Machine Learning Research · doi:10.5555/1953048.2078195 · Cited by 8,539
(2006) · Springer eBooks · Cited by 8,123
Modeling TCP throughput: a simple model and its empirical validation
Padhye, J., Firoiu, V., Towsley, D., & Kurose, J. (1998) · Proceedings of the ACM SIGCOMM '98 conference on Applications, technologies, architectures, and protocols for computer communication · doi:10.1145/285237.285291 · Cited by 2,106 · cited 1×
“Throughput prediction: Prior work either developed approximate analytical models of TCP throughput as a function of packet loss and delay [24, 38] or leveraged Time-series models (e.”
MethodVivaldi: a decentralized network coordinate system
Dabek, F., Cox, R., Kaashoek, F., & Morris, R. (2004) · Proceedings of the 2004 conference on Applications, technologies, architectures, and protocols for computer communications · doi:10.1145/1015467.1015471 · Cited by 1,154 · cited 1×
“Seen in a broader context, CS2P can be regarded as a middle ground between centralized video control platforms (e.g., C3 [23], CFA [29]) and decentralized player-based bi-trate adaptation (e.g., BB [27], FESTIVE [30], MPC [47]).”
BackgroundA Control-Theoretic Approach for Dynamic Adaptive Video Streaming over HTTP
Yin, X., Jindal, A., Sekar, V., & Sinopoli, B. (2015) · Proceedings of the 2015 ACM Conference on Special Interest Group on Data Communication · doi:10.1145/2785956.2787486 · Cited by 1,016 · cited 6×
“…adaptation: While it is possible to develop adaptation approaches without using through-put estimation (e.g., using only the playback buffer occupancy [27]), recent work [47] argues that throughput-aware bitrate adaptation can deliver a better QoE than pure buffer-occupancy based approaches.”
BackgroundMethodInfluentialSelf-citationA buffer-based approach to rate adaptation: evidence from a large video streaming service
Huang, T.-Y., Johari, R., McKeown, N., Trunnell, M., & Watson, M. (2014) · Proceedings of the 2014 ACM conference on SIGCOMM · doi:10.1145/2619239.2626296 · Cited by 1,004
Generalized Boosted Models: A guide to the gbm package
Ridgeway, G. K. (2006) · Cited by 814
Understanding the impact of video quality on user engagement
Dobrian, F., Sekar, V., Awan, A., Stoica, I., Joseph, D., Ganjam, A., Zhan, J., & Zhang, H. (2011) · Proceedings of the ACM SIGCOMM 2011 conference · doi:10.1145/2018436.2018478 · Cited by 673 · cited 2×
“Delivering good application-level video quality-of-experience (QoE) entails new metrics such as low buffering or smooth bitrate delivery [5, 22].”
BackgroundMethodSelf-citationOn the constancy of internet path properties
Zhang, Y., & Duffield, N. (2001) · Proceedings of the First ACM SIGCOMM Workshop on Internet Measurement - IMW '01 · doi:10.1145/505202.505228 · Cited by 466 · cited 1×
“Seen in a broader context, CS2P can be regarded as a middle ground between centralized video control platforms (e.g., C3 [23], CFA [29]) and decentralized player-based bi-trate adaptation (e.g., BB [27], FESTIVE [30], MPC [47]).”
BackgroundImproving Fairness, Efficiency, and Stability in HTTP-Based Adaptive Video Streaming With Festive
Jiang, J., Sekar, V., & Zhang, H. (2014) · IEEE/ACM Transactions on Networking · doi:10.1109/tnet.2013.2291681 · Cited by 384
Self-citationDeveloping a predictive model of quality of experience for internet video
Balachandran, A., Sekar, V., Akella, A., Seshan, S., Stoica, I., & Zhang, H. (2013) · Proceedings of the ACM SIGCOMM 2013 conference on SIGCOMM · doi:10.1145/2486001.2486025 · Cited by 349 · cited 1×
“The result shows that when the error is ≤ 20%, the n-QoE of MPC is close to optimal (>85%).”
BackgroundSelf-citationBroadband internet performance: a view from the gateway
Sundaresan, S., de Donato, W., Feamster, N., Teixeira, R., Crawford, S., & Pescapè, A. (2011) · Proceedings of the ACM SIGCOMM 2011 conference · doi:10.1145/2018436.2018452 · Cited by 304 · cited 1×
“Despite the rich measurement literature in characterizing various Internet path properties (e.g., [21,26,43]), our understanding of throughput variability and predictability is quite limited.”
BackgroundTowards agile and smooth video adaptation in dynamic HTTP streaming
Tian, G., & Liu, Y. (2012) · Proceedings of the 8th international conference on Emerging networking experiments and technologies · doi:10.1145/2413176.2413190 · Cited by 288 · cited 1×
“• Sessions sharing similar critical characteristics tend to exhibit similar throughput patterns.”
BackgroundVideo stream quality impacts viewer behavior: inferring causality using quasi-experimental designs
Krishnan, S. S., & Sitaraman, R. K. (2012) · Proceedings of the 2012 Internet Measurement Conference · doi:10.1145/2398776.2398799 · Cited by 239 · cited 1×
“Even though the above discussion and prior work make the case for throughput prediction, there has been little work on understanding how predictable throughput is in the wild or what types of prediction algorithms we need to use in the context of video bitrate adaptation.”
BackgroundFeedback control for adaptive live video streaming
De Cicco, L., Mascolo, S., & Palmisano, V. (2011) · Proceedings of the second annual ACM conference on Multimedia systems · doi:10.1145/1943552.1943573 · Cited by 226 · cited 2×
“The initial throughput and midstream evolution models can then be plugged into the bitrate selection and adaptation algorithms running either in the video players [30,47] or content delivery servers [14,20].”
MethodLocating internet bottlenecks: algorithms, measurements, and implications
Hu, N., Li, L. (., Mao, Z. M., Steenkiste, P., & Wang, J. (2004) · Proceedings of the 2004 conference on Applications, technologies, architectures, and protocols for computer communications · doi:10.1145/1015467.1015474 · Cited by 201 · cited 1×
“Despite the rich measurement literature in characterizing various Internet path properties (e.g., [21,26,43]), our understanding of throughput variability and predictability is quite limited.”
BackgroundServer-based traffic shaping for stabilizing oscillating adaptive streaming players
Akhshabi, S., Anantakrishnan, L., Dovrolis, C., & Begen, A. C. (2013) · Proceeding of the 23rd ACM Workshop on Network and Operating Systems Support for Digital Audio and Video · doi:10.1145/2460782.2460786 · Cited by 169
On the predictability of large transfer TCP throughput
He, Q., Dovrolis, C., & Ammar, M. (2005) · Proceedings of the 2005 conference on Applications, technologies, architectures, and protocols for computer communications · doi:10.1145/1080091.1080110 · Cited by 161 · cited 1×
“Basics of HTTP-based bitrate adaptation: In HTTP-based video delivery, videos are typically segmented into chunks and each chunk is encoded at different bitrate levels.”
MethodAnalyzing stability in wide-area network performance
Balakrishnan, H., Stemm, M., Seshan, S., & Katz, R. H. (1997) · ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS) · doi:10.1145/258612.258631 · Cited by 148 · cited 1×
“1There has been surprisingly little work and the closest efforts we are aware of are dated and limited in scope [17,49].”
BackgroundCan Accurate Predictions Improve Video Streaming in Cellular Networks?
Zou, X. K., Erman, J., Gopalakrishnan, V., Halepovic, E., Jana, R., Jin, X., Rexford, J., & Sinha, R. K. (2015) · Proceedings of the 16th International Workshop on Mobile Computing Systems and Applications · doi:10.1145/2699343.2699359 · Cited by 142 · cited 4×
“Prior work argues that accurate throughput prediction could help improve initial bitrate selection and the midstream adaptation [45, 47, 50], but fails to provide a concrete roadmap to achieve these benefits.”
BackgroundInfluentialGlasnost: Enabling End Users to Detect Traffic Differentiation
Dischinger, M., Gummadi, K. P., Marcon, M., Mahajan, R., Guha, S., & Saroiu, S. (2010) · Cited by 137
Joshi, S. K., Codreanu, M., & Latva‐aho, M. (2014) · EURASIP Journal on Wireless Communications and Networking · doi:10.1186/1687-1499-2014-1 · Cited by 118
Hidden Markov modeling for network communication channels
Salamatian, K., & Vaton, S. (2001) · ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS) · doi:10.1145/378420.378439 · Cited by 113 · cited 1×
“HMM is an efficient state-based model and has been widely used to predict path and traffic properties [42, 44, 46].”
MethodA machine learning approach to TCP throughput prediction
Mirza, M., Sommers, J., Barford, P., & Zhu, X. (2007) · Proceedings of the 2007 ACM SIGMETRICS international conference on Measurement and modeling of computer systems · doi:10.1145/1254882.1254894 · Cited by 109
Mobile Network Performance from User Devices: A Longitudinal, Multidimensional Analysis
Nikravesh, A., Choffnes, D., Katz-Bassett, E., Mao, Z. M., & Welsh, M. (2014) · Lecture notes in computer science · doi:10.1007/978-3-319-04918-2_2 · Cited by 106 · cited 1×
“, [36]), intra-session RTT variation (e.”
C3: Internet-Scale Control Plane for Video Quality Optimization
Ganjam, A., Jiang, J., Liu, X., Sekar, V., Siddiqi, F., Stoica, I. G., Zhan, J., & Zhang, H. (2018) · KiltHub Repository · doi:10.1184/r1/6468473 · Cited by 86
Self-citationToward a Principled Framework to Design Dynamic Adaptive Streaming Algorithms over HTTP
Yin, X., Sekar, V., & Sinopoli, B. (2014) · Proceedings of the 13th ACM Workshop on Hot Topics in Networks · doi:10.1145/2670518.2673877 · Cited by 82 · cited 1×
“Recent work has shown that accurate throughput prediction can significantly improve the QoE for adaptive video streaming (e.g., [47, 48, 50]).”
BackgroundSelf-citationContinuous-time hidden Markov models for network performance evaluation
Wei, W., Wang, B., & Towsley, D. (2002) · Performance Evaluation · doi:10.1016/s0166-5316(02)00122-0 · Cited by 78
Analyzing the potential benefits of CDN augmentation strategies for internet video workloads
Balachandran, A., Sekar, V., Akella, A., & Seshan, S. (2013) · ACM Internet Measurement Conference (IMC) · doi:10.1145/2504730.2504743 · Cited by 71 · cited 1×
“The result shows that when the error is ≤ 20%, the n-QoE of MPC is close to optimal (>85%).”
BackgroundSelf-citationHow green is multipath TCP for mobile devices?
Lim, Y.-S., Chen, Y., Nahum, E., Towsley, D., & Gibbens, R. (2014) · doi:10.1145/2627585.2627596 · Cited by 60 · cited 2×
“, Holt-Winters [32] and AutoRegressive [24]) and machine-learning models (e.”
BackgroundA measurement study of internet bottlenecks
Hu, N., Li, L., Mao, Z. M., Steenkiste, P., & Wang, J. (2005) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infcom.2005.1498450 · Cited by 52 · cited 1×
“Seen in a broader context, CS2P can be regarded as a middle ground between centralized video control platforms (e.g., C3 [23], CFA [29]) and decentralized player-based bi-trate adaptation (e.g., BB [27], FESTIVE [30], MPC [47]).”
BackgroundA machine learning framework for TCP round-trip time estimation
Arouche Nunes, B. A., Veenstra, K., Ballenthin, W., Lukin, S., & Obraczka, K. (2014) · EURASIP Journal on Wireless Communications and Networking · doi:10.1186/1687-1499-2014-47 · Cited by 43 · cited 1×
“, [37]), and spatial similarity of network performance (e.”
BackgroundApplication-specific path switching
Tao, S., & Guérin, R. A. (2004) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/1027527.1027553 · Cited by 27 · cited 1×
“HMM is an efficient state-based model and has been widely used to predict path and traffic properties [42, 44, 46].”
MethodAnalyzing stability in wide-area network performance
Balakrishnan, H., Stemm, M., Seshan, S., & Katz, R. H. (1997) · ACM SIGMETRICS Performance Evaluation Review · doi:10.1145/258623.258631 · Cited by 26
End-to-end estimation of the available bandwidth variation range
Jain, M., & Dovrolis, C. (2005) · ACM SIGMETRICS Performance Evaluation Review · doi:10.1145/1071690.1064242 · Cited by 11
On the treeness of internet latency and bandwidth
Ramasubramanian, V., Malkhi, D., Kühn, F., Balakrishnan, M., Gupta, A., & Akella, A. (2009) · ACM SIGMETRICS Performance Evaluation Review · doi:10.1145/2492101.1555357 · Cited by 7
Research works printed but not linked: 3
- J. Jiang, V. Sekar, H. Milner, D. Shepherd, I. Stoica, and H. Zhang. CFA: A Practical Prediction System for Video QoE Optimization. In Proc. USENIX NSDI, 2016.
- H. V. Madhyastha, T. Isdal, M. Piatek, C. Dixon, T. Anderson, A. Krishnamurthy, and A. Venkataramani. iPlane: An Information Plane for Distributed Services. In Proc. USENIX OSDI, 2006.
- K. Murphy and M. Dunham. PMTK: Probabilistic Modeling Toolkit. In Proc. NIPS, 2008.
Cited by
Neural Adaptive Video Streaming with Pensieve
Mao, H., Netravali, R., & Alizadeh, M. (2017) · Proceedings of the Conference of the ACM Special Interest Group on Data Communication · doi:10.1145/3098822.3098843 · Cited by 1,357
“However, selecting the right bitrate can be very challenging due to (1) the variability of network throughput [18, 41, 48, 51, 52]; (2) the conflicting video QoE requirements (high bitrate, minimal rebuffering, smoothness, etc.”
BackgroundInfluentialResource Management with Deep Reinforcement Learning
Mao, H., Alizadeh, M., Menache, I., & Kandula, S. (2016) · Proceedings of the 15th ACM Workshop on Hot Topics in Networks · doi:10.1145/3005745.3005750 · Cited by 1,353
“[39], and operate well for different codecs, screen sizes and available bandwidths (e.g., DSL vs. T1).”
BackgroundBoutaba, R., Salahuddin, M. A., Limam, N., Ayoubi, S., Shahriar, N., Estrada-Solano, F., & Caicedo, O. M. (2018) · Journal of Internet Services and Applications · doi:10.1186/s13174-018-0087-2 · Cited by 1,036
Edge Assisted Real-time Object Detection for Mobile Augmented Reality
Liu, L., Li, H., & Gruteser, M. (2019) · The 25th Annual International Conference on Mobile Computing and Networking · doi:10.1145/3300061.3300116 · Cited by 552
A Survey on Bitrate Adaptation Schemes for Streaming Media Over HTTP
Bentaleb, A., Taani, B., Begen, A. C., Timmerer, C., & Zimmermann, R. (2019) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2018.2862938 · Cited by 494
“All theses studies as well as C3 [156], CFA [157], CS2P [158], Pytheas [159] share a common characteristic, which is that there exists a central controller to control, manage and monitor HAS traffic.”
Background
Metadata
- Authors
- Sun, Y., Yin, X., Jiang, J., Sekar, V., Lin, F., Wang, N., Liu, T., & Sinopoli, B.
- Affiliations
- ICT/CAS
- Capital Medical University (CN)
- iQIYI (China) (CN)
- Venue
- Proceedings of the 2016 ACM SIGCOMM Conference
- Publisher
- ACM
- Published
- 2016-08-22
- Pages
- 272-285
- OpenAlex
- W2482797934
- Cited by
- 407
- References
- 38
- Citations by year
- 2026: 20 · 2025: 25 · 2024: 40 · 2023: 43 · 2022: 40 · 2021: 57 · 2020: 55 · 2019: 50 · 2018: 54 · 2017: 21 · 2016: 2
- FWCI
- 36.18
- Type
- conference paper
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- ACM SIGCOMM Conference (SIGCOMM) (publishedVersion)
- Funders
- National Science Foundation (1345305)
- National Natural Science Foundation of China (2012CB315802, 61379133, 61133015)
- National Key Research and Development Program of China (2012CB315802)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › Network Traffic and Congestion Control
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image and Video Quality Assessment
- Physical Sciences › Computer Science › Computer Networks and Communications › Caching and Content Delivery
- Keywords
- throughput prediction, bitrate adaptation, video streaming, quality of experience, hidden Markov models, session clustering
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.