Application, Network and Link Layer Measurements of Streaming Video over a Wireless Campus Network Passive & Active Measurement Workshop 05 Boston, MA, April 1 st, 2005 Feng Li, Jae Chung, Mingzhe Li, Huahui Wu, Mark Claypool, Bob Kinicki Computer Science Department Worcester Polytechnic Institute Worcester, MA, USA
PAM – April 1 st, Outline Introduction Experimental Methods –Tools and Setup –Experimental Design Results and Analysis Conclusions and Future work
PAM – April 1 st, Motivation Increasingly, deployment of streaming multimedia over wireless LANs –Hardware price decreasing. –Wireless link capacity increasing: 11Mbps(802.11b), 54Mbps(802.11g). –Streaming techniques becoming mature.
PAM – April 1 st, Objectives Correlate performance for –Wireless Link Layer –Network Layer –Application Layer Characterize –Streaming Video Parameters
PAM – April 1 st, Outline Introduction Experimental Methods –Tools and Setup –Experimental Design Results and Analysis Conclusions and Future work
PAM – April 1 st, WPI Campus Network LayerToolsPerformance Measures ApplicationMedia TrackerFrame rate, Frame Lost, Encoding bit rate NetworkUDP Ping, WgetRound-Trip time, Packet loss rate, Throughput WirelessTypeperf, WRAPISignal Strength, Frame Retries, Capacity.
PAM – April 1 st, Experimental Parameters Three access points (AP) –Three AP’s on three different floors in the WPI CS Department. Three different reception locations for each AP. –Good, Fair, Bad based on Windows Wireless signal strength indicator. Experiment period –Winter Break, Dec 23 – 25, Dec 28, 29, 2004
PAM – April 1 st, Experimental Methods Each Experiment instance consists of –1 Initial TCP bulk download –8 Videos 2 Clips: High Motion and Low Motion 2 Encoding method –Single Level at 2.5M bps –Multiple Level, 11 levels, max 2.5 M bps. 2 Protocols: TCP and UDP –1 Final TCP bulk download Total: 360 video streaming 360 = 3 APs * 3 Locations * 8 videos * 5 repetitions
PAM – April 1 st, Outline Introduction Experimental Methods –Tools and Setup –Experimental Design Results and Analysis Conclusions and Future work
PAM – April 1 st, High Level Analysis TCP Streaming UDP StreamingTotal Multiple Level Video Single Level Video Subtotal Data Collected 10 data sets were removed from the 360 video runs due to wireless connection failure. TCP Streaming UDP StreamingTotal Multiple Level Video Single Level Video Subtotal
PAM – April 1 st, Signal Strength Analysis Fig.2. Fig.3.
PAM – April 1 st, Wireless Retry Fraction Figure 8 (modified) : Wireless Retry Fraction for Upstream Traffic
PAM – April 1 st, TCP vs UDP Analysis: Frame Rate Figure 6: Frame Rate for TCP and UDP Streaming
PAM – April 1 st, TCP vs UDP Analysis : Round Trip Time Figure 10: Network Round Trip Time for TCP and UDP Streaming
PAM – April 1 st, TCP vs UDP Analysis: Duration Figure 11: Normalized Playout Duration for TCP and UDP Streaming
PAM – April 1 st, Multiple vs Single Analysis: Frame Rate Figure 4: Frame Rate for Multiple and Single Level Encoding
PAM – April 1 st, Multiple vs Single Analysis: Encoding Rate Fig.14 Encoding Bit Rate vs Wireless Link Capacity
PAM – April 1 st, Conclusions At Good wireless reception locations: –Nearly all the video clips played out at a high Frame Rate. –The treatment choices for streaming a video of multiple or single encoding levels and TCP or UDP protocols do not significantly impact performance.
PAM – April 1 st, Conclusions At Bad wireless reception locations: –Multiple level video streams adapt better than single level streams. –Videos streamed using TCP play out at a higher average frame rate than the same video streamed using UDP. –The play out duration for TCP videos is significantly longer than UDP videos.
PAM – April 1 st, Future work To develop new application layer techniques to identify and adapt to challenging wireless transmission situations. To study the behavior at the AP and the interaction of concurrent clients. To concurrently capture burst loss behavior at multiple protocol levels. To evaluate other commercial streaming products: Realplayer TM and Quick time TM player.
Application, Network and Link Layer Measurements of Streaming Video over a Wireless Campus Network Passive & Active Measurement Workshop 05 Feng Li, Jae Chung, Mingzhe Li, Huahui Wu, Mark Claypool, Bob Kinicki CS Department at Worcester Polytechnic Institute Worcester, MA, USA Thank You!
PAM – April 1 st, Appendix: Encoding Bit Rate vs Wireless Link Capacity
PAM – April 1 st, TCP vs UDP Analysis: Network Loss Rate Figure 9: Network Loss Rate for TCP and UDP Stream