performance analysis of home streaming video using orb
DESCRIPTION
Performance Analysis of Home Streaming Video Using Orb. Presented By: Rabin Karki 27 May, 2010. Rabin Karki, Thangam Seenivasan, Mark Claypool and Robert Kinicki Worcester Polytechnic Institute. Introduction. Video streaming largest fraction of Web-based traffic to homes [cite] - PowerPoint PPT PresentationTRANSCRIPT
Performance Analysis of Home Streaming Video Using Orb
Rabin Karki, Thangam Seenivasan, Mark Claypool and Robert Kinicki
Worcester Polytechnic Institute
Presented By:Rabin Karki27 May, 2010
2
Introduction• Video streaming largest fraction of Web-
based traffic to homes [cite]• New trend - users streaming video from
home to Internet• Needs:
–Understand how available bandwidth determined–Measure bandwidth use of new systems–Ascertain video quality
3
Introduction• Orb – free video streaming system from
home to the Internet• Features
– MyCast service– Instant access to photos, music, videos,
television, and other digital content on PC– Access anytime and from any Internet-
connected device• Launched in 2005, now 7+ million users
4
Overview
• Introduction• Goals• Experiments• Results• Conclusions
5
Goals
• Ascertain how Orb determines bandwidth available for streaming
• Measure Orb network traffic under different bandwidth constraints
• Investigate video performance at streaming client
• Understand resource usage at streaming host
6
Overview
• Introduction• Goals• Experiments• Results• Conclusions
7
Orb – Streaming Modes
8
Experiments – Setup
Orb Server
WPI LAN
Host PC Router Client PC
Internet
Direct streaming
Uplink Bandwidth estimation
Downlink Bandwidth estimation
Direct streaming
Host and Client PC• Windows XP running OrbRouter • Linux with NetemNetwork• Direct streamingTools• Wireshark and MediaTracker
9
Experiments – Videos Used
Encode:• Windows Streaming Media, low quality
320x240, 768 kbps, 25 fps• Windows Streaming Media, high quality
1280x720, 1546 kbps, 25 fps• Flash Video
320x214, 320 kbps
Source:• Documentary, High
def, .mov video, 150 seconds
10
Overview
• Introduction• Goals• Experiments• Results• Conclusions
11
Low Quality Video: Frame Rate
• Lower frame rate suggests coarse scaling
• 250 ends later
12
Low Quality Video: Bit rate
• Different encoding levels suggest quality scaling.
• Extremely low bitrate at 250 kbps
13
Low Quality Video: Bandwidth
• Video streamed just below available b/w (except 250 kbps)
14
High Quality Video: Frame rate
• Frame rates similar to low quality video.
15
High Quality Video: Bit rate
• Bitrates different than low quality
• When buffer progress is 100% for some time, bit rate is doubled.
• If buffer progress doesn’t improve, bit rate is reduced.
16
High Quality Video: Bandwidth
• Bandwidth used more closely follows the bandwidth settings than do the encoded bitrates.
17
Host Load
To FLV To WMV0
20
40
60
80
100
120
140
160
180
FLVWMVWMV (HQ)
CPU
Loa
d (%
) Source Format
18
Overview
• Introduction• Goals• Experiments• Results• Conclusions
19
Conclusions
• Bit rate adapts to capacity constraints using TCP
• Temporal and quality scaling– Temporal scaling coarse– Quality scaling smoother for low-quality
video
• Transcoding in real-time– Resource intensive for streaming host
20
Future work
• Other devices• Indirect streaming• Other network settings
21
Thank you!
Questions?