NetworkAdaptive Video Streaming over Wireless Mesh Networks

1
Network-Adaptive Video Streaming over Wireless
Mesh Networks
Lab Seminar

• SangHoon Park
• October 25th, 2007
• Networked Media Laboratory, Department of
  Information and Communication
• School of Information Mechatronics
• Gwangju Institute of Science Technology (GIST)
• shpark_at_nm.gist.ac.kr
• http//nm.gist.ac.kr/shpark

2
Presentation Outline

• Introduction Motivation
• Problem Description
• Proposed System Architecture Scheme
• Implementation
• Experimental Results
• Conclusions Future Work

3
Video Streaming over Wireless Mesh Networks (WMNs)

• WMNs
• A cheap and efficient method for providing
  network connectivity
• Providing real-time multimedia service over WMNs
• VoD (Video on Demand) or Video broadcasting
  services in WMNs-based ubiquitous environment
• Video receivers Multimedia communication with
  Internet servers

ltVideo service infrastructure in Wireless Mesh
Networksgt
VoD Server
Video broadcasting server
Video receivers
4
Challenges in Video Streaming over WMNs

• Higher bit error rate (BER) than that in
  wired-line links
• Packet losses caused by many reasons
• Congestion, Random channel error, Route
  change/break,
• Scarce and time-varying network available
  bandwidth
• Dynamic channel capacity due to various kinds of
  interference
• As increasing hop-count, end-to-end throughput is
  severely degraded
• Lack of QoS support mechanism
• e.g., IEEE 802.11 has serious deficiencies in
  multi-hop environment due to hidden terminal
  effects and contention from neighbor traffic

lt Link-throughput gt
lt End-to-end throughput gt
5
Considerable Solution Approaches

• To handle the impact of channel error
• Efficient and resilient video coding and
  protection (e.g., FEC, delay-constrained ARQ,
  link-layer retransmission, )
• To handle scarce and dynamic network available
  bandwidth
• Scalable video transmission using scalable video
  coding
• Network-adaptive video rate control using network
  monitoring
• Multi-path video transmission
• QoS-supporting in layer
• MAC-layer service differentiation (e.g., IEEE
  802.11e)
• Cross-layer approach (Cross-layer optimization or
  interaction)
• Jointly consider different layers, including
  multimedia application, routing and transporting
  protocol, link layer scheduling, and physical
  layer power control
• We are focusing on the problem Scarce and
  dynamic network available bandwidth
• How to effectively dynamically adapt video
  stream?
• End-to-end video quality improvement

6
Problem Description

• Basic Assumptions
• A video flow can be transmitted in scalable
  fashion (e.g., temporal scalability)
• Base layer l1, enhancement layers l2, l3, , ln
• After video rate adaptation, k video layers are
  transmitted
• A video flow use a single path in WMNs

N5
N4
S
R
N1
N2
N3
N7
N6
7
Problem Description (Cont.)

• Interference due to competing flow
• Network available bandwidth for video flow is
  fluctuating
• Arbitrary Intra-/Inter-background flow
• For the given assumptions,
• How to adapt scalable video according to
  time-varying network available bandwidth to
  improve end-to-end video quality?
• Assumption There is no greedy background flow.
  To cover this issue, multiple flow scheduling
  algorithm or congestion control is required

video flow
background flow
B1
time-varying network available bandwidth
S
R
N1
N2
N3
Be
8
Video Adaptation End-to-End Vs Hop-by-Hop Rate
Adaptation

• End-to-end Video Rate Adaptation
• End-to-end statistics (e.g., loss rate, quality)
  monitoring at receiver
• Sender adapt video based on feedback from
  receiver
• Main drawbacks
• Reliability of feedback will be decreased as the
  congestion is increased
• To guarantee reliable feedback, an additional
  back channel for feedback is needed
• Delay of feedback makes that video adaptation
  reacts slowly to time-varying channel condition

9
Video Adaptation End-to-End Vs Hop-by-Hop Rate
Adaptation

• Hop-by-hop Video Rate Adaptation
• Link statistics monitoring (e.g., MAC-layer loss
  rate) at each intermediate hop
• Intermediate hop adapt video based on own
  monitoring information
• Advantages
• The problems raised in the end-to-end approach
  can be solved
• Overhead Challenges
• Cross-layer design is required
• Monitoring rate adaptation module should be
  deployed at each intermediate hop
• Prioritized packetization at sender is needed

10
Proposed System Architecture Preliminary Design
Intermediate node QoS control
WMN gateway
video streams
Internet
Video Server
Video receiver
Access point
Mesh routers
lt WLANgt
lt WMN Backbonegt
Gateway architecture
Video receiver architecture
Intermediate node architecture
Real-time parsing prioritized packetization
Packet dropping for rate adaptation
playout
Playout buffering
Packet discarding
Wireless channel monitoring
Network monitoring (multiple flow)
Cross-layer design
11
Prioritized Packetization for Temporal Scalability

• We assume
• Profile MPEG-2 TS over RTP
• Priority field
• Each RTP packet contain additional fields in
  application layer
• Frame indexing (may not need)
• Priority (layering information)

ltSimple packetizationgt NOL number of layers
12
Experimental Results WMN Testbed
static routing path for throughput test

• Deployed in GIST DIC 2nd floor
• 1 Gateway (N1), 6 Intermediate nodes (N2N7)
• IEEE 802.11a-based
• Single Interface

KOREN
N1 (gateway)
N2
N3
N4
lt Base-line throughput test (static routing)
N1-gtN6gt
End-to-end throughput
N5
Link-throughput
N6
N7
13
Experimental Setup
video traffic
background traffic

• Video format
• MPEG-2 TS
• Resolution 30fps 720x480
• Bitrate 4Mbps CBR
• GOP IBBPBB
• Interference by background traffic
• N3 -gt N2 24Mbps Pareto UDP traffic
• N4 -gt N3 24Mbps Pareto UDP traffic
• N5 -gt N4 24Mbps Pareto UDP traffic
• N6 -gt N5 24Mbps Pareto UDP traffic
• N7 -gt N6 24Mbps Pareto UDP traffic
• Single video streaming service
• N1 -gt N2 -gt N3 -gt N4 -gt N5 -gt N6 -gt video
  receiver using static routing

KOREN
N1 (gateway)
N2
N3
N4
N5
N6
N7
video receiver
14
Experimental Video Specification

• Temporal scalability of experimental video stream
• Clip length 2 minutes
• GOP IBBPBB, 30fps, 4Mbps
• 4 Temporal layers (l1, l2, l3, l4)
• Rate profile of each temporal layer
• l1 1.52Mbps, l2 0.86Mbps, l3 0.8Mbps, l4
  0.8Mbps
• Frame rate profile of each temporal layer
• l1 5fps, l2 5fps, l3 10fps, l4 10fps

lt Temporal layering of experimental video streamgt
Original GOP
I0
l1 Base layer
P3
l2
B1
B5
Enhancement layers
l3
B2
B6
l4
lt Source video stream traffic characteristic gt
15
Experimental Results Preliminary Results

• Experiment 1 10 times
• According to the background traffic load

lt Background traffic characteristic 2M pareto gt
lt Background traffic pattern gt
16
Experimental Results

• Methodology for end-to-end video quality
  measurement
• Receiving stage Received packet ratio per frame,
  Discarded frame ratio
• Displayed Frame rate
• Discontinuity, Variance of Discontinuity (VoD),
  Autocorrelation of Discontinuity (AoD)

Demuxing
Decoding
Rendering
Receiving
lt th
Y
Frame discarding
17
Preliminary Experimental Results N4-gtN3
background flow
18
Preliminary Experimental Results with No
Background
19
Conclusions Future Work

• Network-Adaptive
• End-to-End Vs Hop-by-hop Video Adaptation in WMNs
• Preliminary Architecture Scheme are proposed
• Future Work
• Performance Evaluation through extensive
  experiments
• Hop-by-hop scheme needed to be improved
• There is still video distortion before video
  adaptation starting due to reactive reponse
• To solve this problem, error recovery and
  adaptive playout can be incorporated (preliminary
  idea!)

20
References
1 J. Jun and M. L. Sichitiu, The nominal
capacity of wireless mesh networks, IEEE
Wireless Communications Magazine, Oct. 2003. 2
Q. Zhang, Video delivery over wireless multp-hop
networks, in Proc. ISPCS, Dec. 2005
21
Thanks ! Q A