daniel farrington nicholas muesch
play

Daniel Farrington, Nicholas Muesch 1 Most online video content is - PowerPoint PPT Presentation

Daniel Farrington, Nicholas Muesch 1 Most online video content is A newer, less popular, growing readily available on demand service What is Live Streaming? What is Video-on-demand? User encodes and uploads Pre-Recorded


  1. Daniel Farrington, Nicholas Muesch 1

  2.  Most online video content is  A newer, less popular, growing readily available on demand service  What is Live Streaming?  What is Video-on-demand? ◦ User encodes and uploads ◦ Pre-Recorded ◦ Server processes ◦ Hosted by Server(s) ◦ Server sends to viewers ◦ Sends to Requested Users VoD Live Streaming 2

  3.  Created in 2005  Created in 2011  Most popular Livestreaming  Second-most popular VoD Website Website ◦ 1.8% downstream traffic (Wallstreet ◦ 18.69 percent of downstream traffic Journal) in NA (CNET.com) YouTube Twitch 3

  4.  YouTube has been thoroughly analyzed ◦ YouTube Traffic Characterization  Network/hardware requirements  Caching content allows for less bandwidth and reduces strain ◦ Understanding characteristics… A YouTube Case Study  Content analysis  E.g. Most common video length is under a minute  Twitch has fewer published papers 4

  5.  Studies such as these can be instrumental in aiding development of future live streaming sites  Room for growth  Our project provides one such study. 5

  6. Te Tech chnica ical Data User r Data Resolutions • Conte tent nt Data Demographics Bitrates • • Preferred content Frame Rates • • Total Streams • Preferred Lengths Key Frame Intervals • • Stream Popularity • Average Use • Game Popularity • Satisfaction • VOD Length • VOD Views • 6

  7.  Used Qualtrics to make survey ◦ Display Blocks  Deployed to CS and IMGD and WPI Grads ◦ Gathered 150+ responses 7

  8.  Automation ◦ Faster, More Reliable, More Data  Scrapy ◦ Python based ◦ Twitch API is RESTful  Batch File and Windows Scheduler ◦ Set a timer ◦ Every hour, runs crawler through cmd 8

  9.  Requests top 100 streams  Grabs stream data  Grabs each channel’s top 100 VoDs  After it reaches the 100 th stream, process repeats 9

  10.  Input 36 channels every day. (Input screenshot)  Channels correlate to most and least popular, 245 each 10

  11.  Survey ◦ Age vs Usage of Twitch and YouTube ◦ Device Usage for Twitch and YouTube ◦ Preferred Viewing Length for Twitch and YouTube videos  Crawler ◦ VoD Length CDF ◦ VoD Viewer CDF ◦ Total Streamers CDF ◦ Game Volatility  Twitch Analyzer ◦ CDF Most and Least Popular Bitrate ◦ CDF Most and Least Popular Streaming Resolutions ◦ Audio Levels of Twitch Broadcasts 11

  12.  Survey ◦ Age vs Usage ge of Twitch tch and YouTube Tube ◦ Device Usage for Twitch and YouTube ◦ Preferred Viewing Length for Twitch and YouTube videos  Crawler ◦ VoD Length CDF ◦ Vo VoD Vi View ewer er CDF ◦ Total Streamers CDF ◦ Ga Game Volat atil ilit ity  Twitch Analyzer ◦ CDF Most and Least Popular Bitrate ◦ CDF Most and d Least Popul ular r Streaming ming Resolu lutions tions ◦ Audio Levels of Twitch Broadcasts 12

  13. 1 0.9 0.8 0.7 le ion Of People 0.6 0.5 Twitch 0.4 YouTube ctio 0.3 Fract 0.2 0.1 0 18-22 23-26 27 or over Age 13

  14. 1 0.9 0.8 on tion 0.7 ributi strib 0.6 e Dist 0.5 tive 0.4 mulati Cumul 0.3 0.2 0.1 0 0 50 100 150 200 250 Views ws 14

  15. Time me (Days) s) 1 6 11 16 21 0 pularity) me popularity) 2 LoL 4 Hearthstone CS: GO ion (Game Dota 2 6 Minecraft 8 stion Poist 10 12 15

  16. 1 0.9 0.8 on 1920x1080 tion 0.7 ributi strib 0.6 e Dist 0.5 tive Most Popular 0.4 mulati Least Popular Cumul 1280x720 0.3 0.2 960x540 0.1 0 0 500000 1000000 1500000 2000000 2500000 Square uare Pixels xels 16

  17.  Twitch is much less known than YouTube ◦ Younger generation knows Twitch better than old ◦ Both young and old generation know YouTube  Most Twitch VoDs are largely unnoticed ◦ 50% of VoDs have less than 8 views  The popularity of games changes unpredictably ◦ E.g. Minecraft fluctuates between 3 rd and 11 th most popular game  There are 2 common resolutions ◦ 1280x720 ◦ 1920x1080 17

  18. Daniel Farrington, Nicholas Muesch 18

Recommend


More recommend