Transcript Document

Towards an Analytical Model for
Characterizing behavior of High-speed VVoIP
Presenter: Prasad Calyam,
OARnet/The Ohio State University, USA
Co-Authors: Andrei Sukhov,
Samara Academy of Transport Engineering, Russia
Warren Daly,
HEANet, Ireland
Alexander Illin
Russian Institute of Public Network (RIPN), Russia
TNC 2005 Presentation, June 9th, 2005
TERENA Networking Conference - 2005
1
Topics of Discussion
Motivations and Study Goals
Terminology
Device factors
Network factors
Human factors
LAN and Internet Testbed Overview
Testbed Testing Results Discussion
Proposed Analytical Model (Work in Progress!)
Conclusion
TERENA Networking Conference - 2005
2
Motivations
Voice and Video over IP (VVoIP) applications
Videoconferencing
H.261, H.263, MPEG-based, …
IPTV for delivering television content over the Internet
Many others…
These demanding applicationsUse high-resolutions (lots of media data!)
Demand real-time handling over the Internet
Need to be able to handle significant User controls
We need to better understand system bottlenecks to
support such VVoIP applications
Analytical model based on characterization of system
parameters could help!
Need to understand performance of the system in end-user
experience terms…
TERENA Networking Conference - 2005
3
Study Goals
We focus on high-speed VVoIP applications
(> 768Kbps) but believe our results are
pertinent at low-speed (< 768 Kbps) also!
Understanding the effects of “device factors” (e.g.
compression schemes, jitter buffering schemes,
packet loss concealment schemes) to maintain
consistent high frame rates on the receiver side
Evaluating the degradation of stream quality due
to “network factors” (e.g. queuing due to
congestion, packet delay, loss and reordering)
Quantifying “human factors” or end-user perceived
experience when there are high expectations of
audiovisual quality in correspondence to the
effects caused by device and network factors
TERENA Networking Conference - 2005
4
Terminology
TERENA Networking Conference - 2005
5
Terminology (2)
UDP1
W1
UDP2
W2
I1
I2
T1
TERENA Networking Conference - 2005
T2
UDP3
W3
UDPn
Wn
In
T3
Tn
Tn+1
6
LAN Testbed
TERENA Networking Conference - 2005
7
Internet Testbed
TERENA Networking Conference - 2005
8
Testing Description
Collecting of VVoIP traces for different dialing
speeds (Kbps)
128, 384, 512, 768, 1024, 1472, 1920
Limited set of end-points, codecs, …
Collect test path network information
Ping, H.323 Beacon, Iperf, Pathping, …
OARnet ↔ HEANet (MOS ≈ 4.33, j < 4 ms, p < 0.12%)
Russia ↔ OARnet/HEANet (MOS<4, j > 40 ms, p > 0.5%)
Examples of “good link” and “bad link” for comparison
Analysis using “Ethereal RTP Analysis Module”
after filtering out the streams
In, Wn, Loss, Re-ordering
TERENA Networking Conference - 2005
9
Ethereal RTP Analysis Module
Steps for analyzing the Traces
Load the packet trace into Ethereal
Trace will contain both forward and reverse direction streams
(Check “Source” and “Destination” IP addresses)
Decode streams as RTP (default is UDP)
This will mark all related packets as belonging to a specific
audio and video codec streams
Analyze individual audio or video streams
Import various information fields as .csv file (“Save as CSV”
option)
Also has wave file generation relating to an audio stream
(“Save Payload” option)
Works only for G.711 Codec streams!
Good for PESQ where you want to compare original and
degraded wave files to obtain Objective MOS information
TERENA Networking Conference - 2005
10
Ethereal RTP Analysis Module (2)
General UDP Stream
decoded as an H.263
payload stream
TERENA Networking Conference - 2005
11
Ethereal RTP Analysis Module (3)
Audio Stream
TERENA Networking Conference - 2005
Video Stream
12
Ethereal RTP Analysis Module (4)
Loss
Re-ordering; 45413 45415 45414
(Observe Sequence #s; could also be 2
consecutive packet losses)
Pink-marked packets relate to either lost or re-ordered packets!
TERENA Networking Conference - 2005
13
Ethereal RTP Analysis Module (6)
An Imported CSV File!
TERENA Networking Conference - 2005
14
Salient Results from Testing
TERENA Networking Conference - 2005
15
Salient Results from Testing (2)
TERENA Networking Conference - 2005
16
Proposed Analytical Model
(Work in Progress!)
Ideally Expected
Value ≈ 4.41
Network Effects
TERENA Networking Conference - 2005
Device Effects
Amelioration by Dejitter
Buffers and Packet loss
Concealment Schemes
17
High-Speed VVoIP Model QMOS Surface
TERENA Networking Conference - 2005
18
Questions?
TERENA Networking Conference - 2005
19