The Hungry Wolf - Bandwidth Measurement of the Congestion Requirements for Video Responsiveness of RealPlayer • Streaming media growing Streaming Video Over UDP – 350,000 hours of online video broadcast ’01 • Voice is 32-64 Kbps, but video has range of data rates Jae Chung, Mark Claypool, Yali Zhu – Videoconference 0.1 Mbps (H.261, MPEG-4) – VCR 1.2 Mbps (MPEG-1) (Technical Report WPI-CS-TR-02-17) Proceedings of the Packet Video Workshop – Broadcast quality 2-4 Mbps (MPEG-2) Nantes, France – HDTV quality 25-34 Mbps (MPEG-2) April 2003 � Thus, potential for more than network capacity http://www.cs.wpi.edu/~claypool/papers/h2h/ The Wolf on the Prowl – Transport The Wolf Pack – Commercial Video Protocols for Video • Streaming video doesn’t like TCP • Commercial products have major impact – Wants rate-based not window-based – Can tolerate some loss • Have been studies characterizing – AIMD causes rate fluctuations commercial traffic � So, use UDP where application controls • But UDP has no congestion control – Bandwidth use, frame rate, user use… • But no work measuring responsiveness, or – Unfair, unfriendly, and even collapse! • Approaches to have router catch lack of it, of commercial video products – Model video as CBR “firehose” (is it?) Specific Motivation Outline 1 Cummulative Density TCP 0.8 • Introduction UDP √ 0.6 • Background ← TCP UDP 0.4 • Experiments • Results 0.2 • Analysis 0 • Conclusions 0 10 20 30 Frame Rate (fps) 1
RealVideo Network Characteristics Outline RTSP • Introduction √ • Background √ Server Data: TCP or UDP • Experiments -Choice, unclear ← -We’ll force • Results • SureStream • Analysis – Allows bandwidth scaling • Buffering • Conclusions – Remove jitter RealVideo Clip Playlist Methodology • If put in controlled environment, may not be representative of clips “in the wild” • Select Real Video Clips • Select large number through search • Construct environment for measuring engines (Yahoo, Google …) • Randomly choose 100 (79 analyzed) congestion response • Construct environment to measure ability • Geography results: to scale (SureStream) – 76 US, 9 Canada, 8 UK, 6 Italy, 1 Germany • Iteratively plan clips varying network – North American dominance likely reflected in typical user locality of reference • Length results: – Median 3 minutes, min 20 seconds, max 30 minutes Responsiveness Measurement Environment Outline Client Router UDP • Introduction √ Hub DSL Internet Client 700Kbps • Background Server √ 10Mbps TCP Token Bucket Filter • Experiments √ • PIII 700 Mhz, 128 MB RAM, Linux 2.4 • RealPlayer version 8.0.3 • Results ← • Sniffing via tcpdump • Analysis • Loss and round-trip time via ping • TBF to limit bandwidth • Conclusions • 2 Measurements for each clip • (Note, RealTracer for MediaScaling) 2
Distribution of Loss Distribution of Latency 1 1 0.8 0.8 Cumulative Density 0.6 Cumulative Density 0.6 075 Kbps 150 Kbps 0.4 075 Kbps 300 Kbps 150 Kbps 0.4 600 Kbps 300 Kbps 600 Kbps 0.2 0.2 0 0 100 200 300 400 500 600 700 800 900 0 RTT (milliseconds) 0 0.05 0.1 0.15 0.2 Loss (fraction) Outline Distribution of Packet Sizes 1 • Introduction √ • Background √ 0.8 • Experiments √ • Results √ Cumulative Density 0.6 • Analysis ← 0.4 – Head to Head – Bandwidth – Scaling 0.2 TCP UDP – Buffering – Smooth • Conclusions 0 0 200 400 600 800 1000 1200 1400 1600 Packet Size (Kbytes) RealPlayer = FairPlayer? RealPlayer = FoulPlayer? 400 600 350 Clip-19 TCP (DSL: BW = 600Kbps, Q = 10Kbytes) Clip-31 TCP (DSL: BW = 600Kbps, Q = 10Kbytes) Clip-19 UDP (DSL: BW = 600Kbps, Q = 10Kbytes) 500 Clip-31 UDP (DSL: BW = 600Kbps, Q = 10Kbytes) 300 400 250 Throughput (Kbps) Throughput (Kbps) 200 300 150 200 100 100 50 0 0 0 50 100 150 200 250 0 50 100 150 200 250 Playout + Buffering Time (Seconds) Playout + Buffering Time (Seconds) 3
Bandwidth Distribution Head-to-Head Bandwidth 1 1 400 75 Kbps 0.9 0.9 150 Kbps 0.8 300 Kbps 0.8 350 600 Kbps 0.7 0.7 Cumulative Density 0.6 600 Kbps Cumulative Density 0.6 0.5 300 0.5 300 Kbps Average Bandwidth for UDP (Kbps) 0.4 0.4 0.3 0.3 250 0.2 Average Bandwidth for TCP 0.2 Average Bandwidth for TCP Average Bandwidth for UDP Average Bandwidth for UDP 0.1 0.1 200 0 0 0 100 200 300 400 500 600 0 50 100 150 200 250 300 Average Bandwidth for 600 Kbps (Kbps) Average Bandwidth for 300 Kbps (Kbps) 1 1 150 0.9 0.9 0.8 0.8 0.7 0.7 100 Cumulative Density Cumulative Density 0.6 0.6 0.5 0.5 75 Kbps 150 Kbps 50 0.4 0.4 0.3 0.3 0 0.2 Average Bandwidth for TCP 0.2 Average Bandwidth for TCP Average Bandwidth for UDP Average Bandwidth for UDP 0 50 100 150 200 250 300 350 400 0.1 0.1 Average Bandwidth for TCP (Kbps) 0 0 0 20 40 60 80 100 120 140 0 10 20 30 40 50 60 70 Average Bandwidth for 150 Kbps (Kbps) Average Bandwidth for 75 Kbps (Kbps) Bandwidth Difference Distribution Mostly TCP-Friendy! 1 • Remove low bandwidth (36%) • Then remove unscalable (14%) 0.8 Cumulative Density 0.6 0.4 0.2 75 Kbps 150 Kbps 300 Kbps 600 Kbps 0 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 Normalized Average Bandwidth Difference (TCP-UDP) Outline Media Scaling Distribution 1 • Introduction √ 0.9 • Background √ 0.8 • Experiments √ 0.7 Cumulative Density • Results √ 0.6 • Analysis 0.5 √ 0.4 – Head to Head √ – Bandwidth 0.3 ← – Scaling 0.2 – Buffering 0.1 – Smooth 0 • Conclusions 1 2 3 4 5 6 7 8 9 Number of Scales (Coded-Bandwidth) 4
Media Scaling Dynamics (1) Media Scales 250000 local bw limit (35 kbps) Scale (Coded-BW) Movement for Clip-65 TCP Scale (Coded-BW) Movement for Clip-65 UDP 200000 Coded-Bandwidth (bps) 150000 100000 50000 0 0 50 100 150 200 250 300 350 Playout + Buffereing Time (sec) Media Scaling Dynamics (2) Media Scaling Distributions 300000 1 local bw limit (35 kbps) Scale (Coded-BW) Movement for Clip-78 TCP 0.9 250000 Scale (Coded-BW) Movement for Clip-78 UDP 0.8 0.7 200000 Coded-Bandwidth (bps) Cumulative Density 0.6 150000 0.5 0.4 100000 0.3 0.2 Number of Scale Changes Seen for TCP 50000 Number of Scale Changes Seen for UDP 0.1 0 0 0 20 40 60 80 100 120 140 0 2 4 6 8 10 Playout + Buffereing Time (sec) Number of Scale (Coded-Bandwidth) Changes Media Scaling Adaptation Speed Outline 1 • Introduction √ 0.9 • Background √ 0.8 • Experiments √ 0.7 • Results Cumulative Density √ 0.6 • Analysis 0.5 √ – Head to Head 0.4 √ – Bandwidth 0.3 √ – Scaling 0.2 Scale Adaptation Speed for TCP Scale Adaptation Speed for UDP ← – Buffering 0.1 – Smooth 0 • Conclusions 0 10 20 30 40 50 60 70 80 Elaps Time in Seconds: 0 to time(Coded-BW < 35kbps) 5
Buffering Rate to Playout Rate Buffering Rate to Playout Rate Distribution 7 1 Average Buffering Rate / Average Steady Playout Rate 6 TCP (All DSL-TBF Runs) UDP (All DSL-TBF Runs) 0.8 5 Cumulative Density 0.6 4 3 0.4 2 0.2 TCP (All LAN Runs) UDP (All LAN Runs) 1 0 0 0 1 2 3 4 5 6 7 0 50 100 150 200 250 300 350 400 Average Buffering Rate / Average Steady Playout Rate Average Steady Playout Rate (Kbps) Smoothness at each Bottleneck Smoothness 1 1 TCP TCP UDP UDP 0.9 0.9 1 0.8 0.8 TCP UDP 0.7 0.7 0.9 Cumulative Density 0.6 Cumulative Density 0.6 0.5 0.5 0.8 0.4 0.4 0.3 0.3 0.7 0.2 0.2 Cumulative Density 0.1 0.6 0.1 0 1/2 1 2 0 Throughput Ratio for 600 Kbps 1/2 1 2 0.5 Throughput Ratio for 300 Kbps 1 1 TCP TCP UDP UDP 0.9 0.9 0.4 0.8 0.8 0.3 0.7 0.7 Cumulative Density Cumulative Density 0.6 0.6 0.2 0.5 0.5 0.4 0.4 0.1 0.3 0.3 0.2 0.2 0 1/2 1 2 0.1 0.1 Throughput Ratio 0 0 1/2 1 2 1/2 1 2 Throughput Ratio for 150 Kbps Throughput Ratio for 75 Kbps Discussion of Results Conclusions • No concrete incentives to respond to congestion • Response to congestion important to – In fact, may be “rewarded” for not • However, clearly responds Internet • Measured responsiveness of RealVideo – often TCP-Friendly • Content providers need to provide chance • RealVideo over UDP is largely responsive for scalability – receives typically same bandwidth as TCP • But buffering at higher rate is bad for – often TCP friendly network • TCP can be as smooth as UDP • So why not TCP? � API is limiting 6
Future Work? Future Work • Other commercial Players – Microsoft Media Player • Live clips (versus pre-recorded clips) • Perceptual quality of video over TCP versus UDP • Characterization of clips on Internet – So can examine “typical” clips 7
Recommend
More recommend
