Mobile Opportunistic Videoondemand in Opportunistic Networks

1
Mobile Opportunistic Video-on-demand in
Opportunistic Networks
2nd NML Seminar, 2007
Latest results and toward MOVi in multi-channel
environments

• 23rd , Nov, 2007
• Hayoung Yoon
• hyyoon_at_nm.gist.ac.kr
• Networked Media Laboratory
• Dept. of Information Communications
• Gwangju Institute of Science Technology

2
Agenda

• Background
• Scalability Issues in VoD services in Wireless
  Networks
• MOVi Architecture
• Latest Simulation and Experiment Results
• MOVi in Multi-channel
• Multi-channel iDLS
• Neighbor Sensing in Multi-channel environments
• Optimal Scheduler Design
• Conclusion Future Work

3
Background

• Video On-Demand (VoD) provides highly flexible
  streaming service to end-users
• Known problems in VoD service are
• Efficiency
• Multiple copies
• Scalability
• Centralized design
• Reliability
• No perfect transport protocol

Contents
T1
T2
4
Scalability Issues in VoD Service
main Server

• System
• Distributed Servers
• A lot of idea in reducing workload at the server
• Bandwidth
• Too much workload in Core network
• Caching Caching_at_ICN02
• CDN Utube_at_Mobisys07
• P2P Approaches Coolstreaming P2cast

Server
Server
CDN
CDN
T1
T2
5
VoD in Wireless Networks

• Access networks are infrastructure wireless
  network
• WCore WAccess
• P2P strategy is not good idea
• Double resource consumption problem
• CDN is limited
• Relatively small bandwidth of wireless networks
• Need a novel way to enhance the utilization of
  wireless network regarding VoD application
  characteristics!

6
MOVi Architecture
Video Contents
Control Server

• Broadcast is NOT flexible
• VoD takes O(N) resources
• Mobility increases capacity in Mobile Networks
  IEEE TIT_at_2002
• Issues
• - Framework (IEEE 802.11-based)
• - Optimal Scheduler
• - Performance Boundary
• - Effective Localized Comm.
• - Mobility Awareness
• - PHY Enhancement

- Wired Networks -
- Wireless Access Networks -
- Mobile Opportunistic Networks -
7
Example Scenario of MOVi
MS
AP
AP
1,2,3,4
5,6
MC(1)
1,2,3,4
3,4,5,6
MC(2)
Higher Rate Lower Power!
5,6
8
MS Scheduler Design

• Update of each MC status
• Who are neighbors?
• Remaining Buffer size
• Sort MCs w.r.t. its Remaining Buffer
• Evaluate CONTACT
• Schedule MC-to-MC communications
• Schedule MS-to-MC communications

9
Simulation Configurations

• Topology
• 8APs (all APs are connected to OCDP through wired
  line)
• 810 OMs
• Mobility Trace from Crawdad dataset Kim_at_INFOCOM
  2006
• Channel
• Riciean Fading model (k6, V2.5m/s)
• PHY
• IEEE 802.11a (8 different bit-rate)
• ARF (auto-rate fallback) at APs
• Distributed Dynamic Rate and Power Control over
  OMCF Lim_at_Mobicom2006
• 40 mW TX pow. For up/downlink
• Set Same Freq. for all APs and OMs for both Conv.
  and OPAN case
• MAC
• EDCA (high priority for controls and the low for
  data TX)

10
Latest Simulation Results
Play Continuity Index MOVi and CVoD Performance
Comparison
11
Latest Simulation Results
of MCs MOVi and CVoD Performance Comparison
12
Latest Simulation Results
Prebuffering time MOVi and CVoD Performance
Comparison
13
Latest Simulation Results
PHY-rate Distribution over TX/RX packets MOVi
and CVoD Performance Comparison
54Mbps
48Mbps
54Mbps
36Mbps
48Mbps
36Mbps
24Mbps
24Mbps
14
Latest Simulation Results
Battery Consumption for TX packets MOVi and
CVoD Performance Comparison
15
Latest Simulation Results
Battery Consumption for RX packets MOVi and
CVoD Performance Comparison
16
Latest Simulation Results
Battery Consumption for RXTX packets MOVi and
CVoD Performance Comparison
17
Experiment Setup

• Each MC request video contents at different time
• No movement during experiments
• 10 times of iterations and get average PCI
  performance

18
Latest Experiment Results
Sorted MOVi and CVoD PCI Performance Comparison
19
Latest Experiment Results
MOVi Experiment and Simulation Validation
20
Latest Experiment Results
CVoD Experiment and Simulation Validation
21
Toward MOVi in multi-channel environments
22
What Must be Considered in Multi-channel
Environment

• Sensing Neighbors
• Geometrically Closed two neighbors may not be on
  the same channel
• MS Scheduling Procedure
• Purpose
• MC-to-MC communication
• Separation of Communication channel
• MC1 and MC2 opportunistically move to the other
  channel for communication
• Cognitive network

FA
FB
N3
N2
N1
54Mbps
12Mbps
N2
N2
N1
N1
N4
N4
N3
N3
23
Issues in Sensing Neighbors

• Assume we have single device
• Passive
• Sensing other channel (spectrum) periodically
  (e.g., for 100msec every 1 sec)
• Loss of bandwidth for sensing time
• Cant detect silent nodes
• Active
• Broadcast packet to other channel periodically
• Loss of bandwidth for broadcasting as well as
  overhead of broadcast
• Detect silent nodes well

24
Issues in Optimal Scheduler Design

• Optimal P2P scheduling is out there
  Scheduling07
• Wired networks, non-saturation case, and normal
  data transfer
• Mixed integer linear programming (MILP)
• 2 nodes and 2 packets case takes 1 day to get the
  optimal solution
• Graph Coloring Problem
• Given number of N channels and MC connection
  graph G(V,E)
• Find Optimal schedule S finishes job as soon as
  possible
• If the number of the connections (MC-to-MC and
  MS-to-MC) are smaller than N the solution is
  trivial
• We will focus on the other case
• Finding Domination Node Problem
• High degree nodes have more possibility to
  contribute to their neighbors
• Tradeoff
• Buffer underflow Play discontinuity
• Unfair battery life time

25
Issues in MC-to-MC communications

• Synchronized channel switching at the both MCs
• How much is the channel frequency switching
  delay?
• How accurate the delay bounds?

26
Conclusion

• MOVi (Formally known as VoD system over OPAN) has
  implemented in a Single device and Single channel
  environment
• Centralized status tracking and decentralized
  data distribution
• Direct MC-to-MC communication
• Pre-buffering level-based centralized scheduling
• Playout quality and number of users MOVi can
  provide is roughly 200 of what CVoD can
• Various issues to enhance MOVi in the
  multi-channel environment