Windows Streaming Media Performance Analysis on a IEEE g Residential Network The Seventh International Conferences on Wireless and Optical Communications Montreal, Canada, May 30 – June 1, 2007 Travis Grant, Saurabh Gupta, Harshal Pandya and Robert Kinicki Worcester Polytechnic Institute Worcester, MA, USA Presenter - Bob Kinicki
WOC 2007 May 31, Outline MotivationMotivation Previous WorkPrevious Work ExperimentsExperiments Results and AnalysisResults and Analysis ConclusionsConclusions
WOC 2007 May 31, Motivation One objective is to anticipate support for HD video on demand (VOD) over residential Wireless Local Area Networks (WLANs) via IPTV and broadband services.One objective is to anticipate support for HD video on demand (VOD) over residential Wireless Local Area Networks (WLANs) via IPTV and broadband services. We investigate a current off-the-shell streaming server’s (Windows Streaming Media) delivery of high quality video over a residential WLAN under the assumption there are no special administrative adjustments by the customer.We investigate a current off-the-shell streaming server’s (Windows Streaming Media) delivery of high quality video over a residential WLAN under the assumption there are no special administrative adjustments by the customer. Experimental measurements can provide insight into streaming application performance behavior when wireless errors are encountered in the g MAC layer.Experimental measurements can provide insight into streaming application performance behavior when wireless errors are encountered in the g MAC layer.
WOC 2007 May 31, Outline MotivationMotivation Previous WorkPrevious Work ExperimentsExperiments Results and AnalysisResults and Analysis ConclusionsConclusions
WOC 2007 May 31, Measurement Studies of [Bai and Williamson 04] “The Effects of Mobility on Wireless Media Streaming Performance”[Bai and Williamson 04] “The Effects of Mobility on Wireless Media Streaming Performance” –Create their own AP device to vary queue size. –Downstream measurements of UDP videos show WLAN supports easily two fixed clients receiving 1Mbps video clips with AP queue less than 30 buffers. –When one client becomes mobile, it goes through “bad” locations and frames get discarded, rate adaptation moves to 1 Mbps, AP queue backlogs and overflows!! –When one client fixed and one client mobile, mobile client kills performance of fixed client because the MAC-layer queue fills with frames from poorly-connected client. The AP queue is the bottleneck. [Yarvis et al. 05] “Characteristics of Wireless Networks”[Yarvis et al. 05] “Characteristics of Wireless Networks” –Consider: transmission rate, transmission power, node location, house type. –Conduct measurements in three homes with link layer retransmissions disabled. –Discover: wireless performance can be quite asymmetric, node placement can be a key factor, no correlation with physical distance.
WOC 2007 May 31, Measurement Studies of [F. Li et al. 05] “Application, Network and Link Layer Measurements of Streaming Video over a Wireless Campus Network”[F. Li et al. 05] “Application, Network and Link Layer Measurements of Streaming Video over a Wireless Campus Network” –Both papers involve clients at good and bad locations. –TCP versus UDP streaming matters at a bad location. –Multi-encoded versus single encoded matters at a bad location. [Gretarsson et al. 05] “Performance Analysis of the Intertwined Effects between Network Layers for g Transmissions”[Gretarsson et al. 05] “Performance Analysis of the Intertwined Effects between Network Layers for g Transmissions” –A single TCP download only managed 18.8 Mbps! –Application behavior impacts WLAN performance of concurrent applications. –AP queuing strategy can impact performance beyond the performance anomaly.
WOC 2007 May 31, Outline MotivationMotivation Previous WorkPrevious Work ExperimentsExperiments Results and AnalysisResults and Analysis ConclusionsConclusions
WOC 2007 May 31, Experimental Configuration
WOC 2007 May 31, Measurement Tools WRAPI and MediaTracker collect data on the client.WRAPI and MediaTracker collect data on the client. See our Web page for free WLAN measurement tools our Web page for free WLAN measurement tools Other data was collected using Etheral on the wireless sniffer.Other data was collected using Etheral on the wireless sniffer.
WOC 2007 May 31, Three Test Videos 1.HD 60 sec, 8.1 Mbps singly-encoded HD sample video from Microsoft Web site. 2.SD 120 sec, 5.1 Mbps CBR video that exceeded SD quality. 3.LowRes 126 sec, 1.1 Mbps video * While streams were sent using both TCP and UDP, only TCP results are reported in this paper.
WOC 2007 May 31, Outline MotivationMotivation Previous WorkPrevious Work ExperimentsExperiments Results and AnalysisResults and Analysis ConclusionsConclusions
WOC 2007 May 31, Good versus Bad Location Signal Strength of Visible APs Only Good results reported in this paper. Received Signal Strength (RSSI) can impact MAC layer rate adaptation algorithms.
WOC 2007 May 31, HD Video CPU Utilization Video stalls
WOC 2007 May 31, HD Video Tests High Definition Video Actual Playback Times on g Frequent Audio/Video stalls occurred!
WOC 2007 May 31, SD and LowRes Videos
WOC 2007 May 31, SD Video Test SD Video streams successfully
WOC 2007 May 31, LowRes Video High Error Counts
WOC 2007 May 31, SD Video Measurements
WOC 2007 May 31, Conclusions and Conjectures HD video did not successfully stream over residential WLAN. Possible reasons are:HD video did not successfully stream over residential WLAN. Possible reasons are: –Restrictions within the WMS server configuration force playout buffer induced stalls. –Only one encoded level for HD video limits Intelligent Streaming adjustments. –MAC layer rate adaptation when MAC frame loss count is high negatively impacts Intelligent Streaming. –MAC layer retries cause problems for application bandwidth estimation probes. 5.1 SD video and LowRes video did successfully stream over WLAN.5.1 SD video and LowRes video did successfully stream over WLAN. {Future Work} What happens when there are other concurrent flows?{Future Work} What happens when there are other concurrent flows?
Windows Streaming Media Performance Analysis on a IEEE g Residential Network Travis Grant, Saurabh Gupta, Harshal Pandya and Robert Kinicki WPI Computer Science Department Worcester, MA, USA Thank You! Questions ??