An official website of the United States government
Low-latency is a critical user Quality-of-Experience (QoE) metric for live video streaming. It poses significant challenges for streaming over the Internet. In this paper, we explore the design space of low-latency live video streaming by developing dynamic models and optimal control strategies. We further develop practical live video streaming algorithms within the Model Predictive Control (MPC) framework, namely MPC-Live, to maximize user QoE by adapting the video bitrate while maintaining low end-to-end video latency in dynamic network environment. Through extensive experiments driven by real network traces, we demonstrate that our live video streaming algorithms can improve the performance dramatically within latency range of two to five seconds.
more »
« less
Learning in situ: a randomized experiment in video streaming
We describe the results of a randomized controlled trial of video-streaming algorithms for bitrate selection and network prediction. Over the last year, we have streamed 38.6 years of video to 63,508 users across the Internet. Sessions are randomized in blinded fashion among algorithms. We found that in this real-world setting, it is difficult for sophisticated or machine-learned control schemes to outperform a "simple" scheme (buffer-based control), notwithstanding good performance in network emulators or simulators. We performed a statistical analysis and found that the heavy-tailed nature of network and user behavior, as well as the challenges of emulating diverse Internet paths during training, present obstacles for learned algorithms in this setting. We then developed an ABR algorithm that robustly outperformed other schemes, by leveraging data from its deployment and limiting the scope of machine learning only to making predictions that can be checked soon after. The system uses supervised learning in situ, with data from the real deployment environment, to train a probabilistic predictor of upcoming chunk transmission times. This module then informs a classical control policy (model predictive control). To support further investigation, we are publishing an archive of data and results each week, and will open our ongoing study to the community. We welcome other researchers to use this platform to develop and validate new algorithms for bitrate selection, network prediction, and congestion control.
more »
« less
- PAR ID:
- 10186616
- Date Published:
- Journal Name:
- 17th USENIX Symposium on Networked Systems Design and Implementation (NSDI '20)
- Page Range / eLocation ID:
- 495-511
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The performance of Adaptive Bitrate (ABR) algorithms for video streaming depends on accurately predicting the download time of video chunks. Existing prediction approaches (i) assume chunk download times are dominated by network throughput; and (ii) apriori cluster sessions (e.g., based on ISP and CDN) and only learn from sessions in the same cluster. We make three contributions. First, through analysis of data from real-world video streaming sessions, we show (i) apriori clustering prevents learning from related clusters; and (ii) factors such as the Time to First Byte (TTFB) are key components of chunk download times but not easily incorporated into existing prediction approaches. Second, we propose Xatu, a new prediction approach that jointly learns a neural network sequence model with an interpretable automatic session clustering method. Xatu learns clustering rules across all sessions it deems relevant, and models sequences with multiple chunk-dependent features (e.g., TTFB) rather than just throughput. Third, evaluations using the above datasets and emulation experiments show that Xatu significantly improves prediction accuracies by 23.8% relative to CS2P (a state-of-the-art predictor). We show Xatu provides substantial performance benefits when integrated with multiple ABR algorithms including MPC (a well studied ABR algorithm), and FuguABR (a recent algorithm using stochastic control) relative to their default predictors (CS2P and a fully connected neural network respectively). Further, Xatu combined with MPC outperforms Pensieve, an ABR based on deep reinforcement learning.more » « less
-
The performance of Adaptive Bitrate (ABR) algorithms for video streaming depends on accurately predicting the download time of video chunks. Existing prediction approaches (i) assume chunk download times are dominated by network throughput; and (ii) apriori cluster sessions (e.g., based on ISP and CDN) and only learn from sessions in the same cluster. We make three contributions. First, through analysis of data from real-world video streaming sessions, we show (i) apriori clustering prevents learning from related clusters; and (ii) factors such as the Time to First Byte (TTFB) are key components of chunk download times but not easily incorporated into existing prediction approaches. Second, we propose Xatu, a new prediction approach that jointly learns a neural network sequence model with an interpretable automatic session clustering method. Xatu learns clustering rules across all sessions it deems relevant, and models sequences with multiple chunk-dependent features (e.g., TTFB) rather than just throughput. Third, evaluations using the above datasets and emulation experiments show that Xatu significantly improves prediction accuracies by 23.8% relative to CS2P (a state-of-the-art predictor). We show Xatu provides substantial performance benefits when integrated with multiple ABR algorithms including MPC (a well studied ABR algorithm), and FuguABR (a recent algorithm using stochastic control) relative to their default predictors (CS2P and a fully connected neural network respectively). Further, Xatu combined with MPC outperforms Pensieve, an ABR based on deep reinforcement learning.more » « less
-
The performance of Internet services—be it file download completion times, video quality, or lag-free video conferencing—is heavily influenced by network parameters. These include the bottleneck bandwidth, network delays, and how fairly the bottleneck link is shared with other services. However, current techniques to evaluate service performance in emulated and simulated networks suffer from three major issues: (a) testing predominantly in settings representing the "edge" of the Internet, and not the core; (b) focus on evaluating Congestion Control Algorithms (CCAs), neglecting the impact of application-level controls like Adaptive-Bitrate (ABR) algorithms on network performance; (c) testing in settings that do not necessarily reflect the network conditions experienced by services with expansive CDNs. The goal of this thesis is to improve the state of the art in emulated testing for a more up-to-date evaluation of Internet service performance. To highlight the need to perform Internet evaluations in settings representing congestion at the core of the Internet, we test CCAs with core Internet speeds and flow counts. We find that this dramatically alters fairness outcomes, and challenges long-standing assumptions about CCA behavior that were built on measurements performed at in settings representing the edge of the Internet, emphasizing the need to run Internet evaluations in more diverse settings. We then challenge the implicit assumption that CCA evaluations alone are suf- ficient to predict the network behavior of services that use them. We perform this analysis through the lens of fairness, and build Prudentia, an Internet fairness watch- dog, that measures how fairly two Internet services can share a bottleneck link. In addition to discovering extreme unfairness on the Internet today, we gain key insights into improving current testing methodology – (a) The most and least fair services both use variants of the same CCA, highlighting the need to test services in addition to CCAs; (b) network settings can drastically affect even service-level fairness outcomes, necessitating their careful selection. Lastly, we infer the network conditions experienced by users of Netflix, a global video streaming provider, and contrast them with those used in typical Internet evaluations. We find that Netflix users experience shorter RTTs, greater maximum observed queuing delay, and greater ACK aggregation, all parameters that play an important role in determining CCA behavior. This highlights the need for more service operators to run similar analyses and share their respective perspectives of prevalent network conditions, so that the networking community can include these settings in the design and evaluation of Internet services.more » « less
-
The adaptive bitrate selection (ABR) mechanism, which decides the bitrate for each video chunk is an important part of video streaming. There has been significant interest in developing Reinforcement-Learning (RL) based ABR algorithms because of their ability to learn efficient bitrate actions based on past data and their demonstrated improvements over wired, 3G and 4G networks. However, the Quality of Experience (QoE), especially video stall time, of state-of-the-art ABR algorithms including the RL-based approaches falls short of expectations over commercial mmWave 5G networks, due to widely and wildly fluctuating throughput. These algorithms find optimal policies for a multi-objective unconstrained problem where the policies inherently depend on the predefined weight parameters of the multiple objectives (e.g., bitrate maximization, stall-time minimization). Our empirical evaluation suggests that such a policy cannot adequately adapt to the high variations of 5G throughput, resulting in long stall times. To address these issues, we formulate the ABR selection problem as a constrained Markov Decision Process where the objective is to maximize the QoE subject to a stall-time constraint. The strength of this formulation is that it helps mitigate the stall time while maintaining high bitrates. We propose COREL, a primal-dual actor-critic RL algorithm, which incorporates an additional critic network to estimate stall time compared to existing RL-based approaches and can tune the optimal dual variable or weight to guide the policy towards minimizing stall time. Our experiment results across various commercial mmWave 5G traces reveal that COREL reduces the average stall time by a factor of 4 and the 95th percentile by a factor of 2.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
