A Quest for an Internet Video Quality-of-Experience Metric Athula Balachandran, Vyas Sekar, Aditya Akella, Srinivasan Seshan, Ion Stoica, Hui Zhang 1 Internet Video is taking off Improve Users’ Quality of Experience 2 Video Quality Metrics: The State of the Art Subjective Scores (e.g., Mean Opinion Score) Objective Score (e.g., Peak Signal to Noise Ratio) 3 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Buffer empty Buffer filled up Switch bitrate 4 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Buffer empty Buffer filled up Switch bitrate 5 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Buffer empty Buffer filled up Switch bitrate 6 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Buffer empty Buffer filled up Switch bitrate 7 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Buffer empty Buffer filled up Switch bitrate 8 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Buffer empty Buffer filled up Switch bitrate 9 Problem 1: New Effects, New Metrics PLAYER Joining STATES Playing Buffering Playing EVENTS Buffer filled up Join Time Buffer empty Buffering Ratio Rate of buffering Buffer filled up Switch bitrate Rate of switching Average bitrate 10 Problem 2: Opinion Scores Engagement Opinion Scores - Not representative of “in the wild” experience - Combinatorial explosion of parameters Engagement as replacement for opinion score. (e.g., Play time, customer return rate) 11 Internet Video QoE Subjective Scores MOS Objective Scores PSNR 12 Internet Video QoE Subjective Scores MOS Engagement (e.g., Fraction of video viewed) Objective Scores PSNR 13 Internet Video QoE Subjective Scores MOS Engagement (e.g., Fraction of video viewed) Objective Scores PSNR Join Time, Avg. bitrate, …? 14 Internet Video QoE Subjective Scores MOS Engagement (e.g., Fraction of video viewed) Objective Scores PSNR Join Time, Avg. bitrate, …? f(Join Time, Avg. bitrate, …) 15 Internet Video QoE Subjective Scores MOS Engagement (e.g., Fraction of video viewed) Objective Scores PSNR Join Time, Avg. bitrate, …? f(Join Time, Avg. bitrate, …) 16 Outline • Need for a unified QoE • What makes this hard? • Our proposed approach 17 Engagement Challenge: Complex Engagement-to-metric Relationships Quality Metric 18 Engagement Engagement Challenge: Complex Engagement-to-metric Relationships Non-monotonic Average bitrate Quality Metric [Dobrian et al. Sigcomm 2011] 19 Engagement Engagement Challenge: Complex Engagement-to-metric Relationships Non-monotonic Quality Metric [Dobrian et al. Sigcomm 2011] Engagement Average bitrate Threshold Rate of switching 20 Challenge: Complex Metric Interdependencies Join Time Rate of buffering Bitrate Rate of switching Buffering Ratio 21 Challenge: Complex Metric Interdependencies Join Time Rate of buffering Bitrate Rate of switching Buffering Ratio 22 Challenge: Complex Metric Interdependencies Join Time Rate of buffering Bitrate Rate of switching Buffering Ratio 23 Challenge: Complex Metric Interdependencies Join Time Rate of buffering Avg. bitrate Rate of switching Buffering Ratio 24 Need to learn these complex engagement-to-metric relationships and metric-to-metric dependencies 25 Casting as a Learning Problem Need to learn these complex engagement-to-metric relationships and metric-to-metric dependencies Engagement Quality Metrics MACHINE LEARNING QoE Model 26 Impact of the ML algorithm • Classify engagement into uniform classes • Accuracy = # of accurate predictions/ # of cases ML algorithm must be expressive enough to handle the complex relationships and interdependencies 27 Challenge: Confounding Factors Live and VOD sessions experience similar quality 28 Challenge: Confounding Factors However, user viewing behavior is very different 29 Challenge: Confounding Factors Devices Connectivity User Interest Need systematic approach to identify and handle confounding factors 30 Domain-specific Refinement Engagement Quality Metrics MACHINE LEARNING QoE Model 31 Domain-specific Refinement Engagement Confounding Factors Quality Metrics MACHINE LEARNING QoE Model 32 Improved prediction accuracy Refined ML models can handle confounding factors 33 Concluding Remarks • Internet Video needs unified quantitative QoE • What makes this hard? – – – Complex engagement-to-metric relationships Complex metric-to-metric interdependencies Confounding factors (e.g., genre, device) • Promising start – Machine learning + domain-specific refinements • Open Challenges – – Coverage over confounding factors System Design 34