Provision of VCR-like Functions in Multicast VoD

1
Provision of VCR-like Functions in Multicast VoD
2
Presentation Map

• Introduction
• Problem identification
• Review on solutions
• Issues in algorithm design
• Q A

3
Introduction

• Targets
• Based on multicast network architecture,
  implement VCR-like functions
• Minimize additional system resources for the
  resultant system

4
Problem Identification

• TVoD (True Video-on-Demand)
• NVoD (Near Video-on-Demand)
• Comparisons server bandwidth, response time,
  play-out point control

• One channel is dedicated to one client
• Waiting time depends on system load
• Clients have total control once they are admitted

L

• Adjacent multicast channels stream video data
  with time lag of Tr
• Clients have to wait for the start of movie
  broadcast, max wait Tr, mean Tr / 2

staggered multicast channels
Tr
Multicast network
5
Problem Identification

• Challenge
• How to achieve both
• Bounded access latency in NVoD
• Total control in TVoD
• with a given set of system restrictions (e.g.
  server bandwidth, access bandwidth)
• Current solutions
• Periodic broadcasting protocols

6
Problem Identification

• What are VCR-like operations?
• Pause, stop
• Fast forward / backward
• Slow motion
• Jump forward / backward
• Result change of client play-out point relative
  to the movie

7
Problem Identification

• VCR-like functions resumption
• Client play-out point and server broadcasting
  point are out of phase
• Client has to either
• Be served by unicast contingency channels ?
  bandwidth save due to multicast is lost, or
• Wait for next broadcast of play-out point ? the
  waiting time is unacceptable under normal system
  configurations (e.g. staggered broadcasting of
  2hr movie with 25 channels, mean wait 144s)

bp ? server broadcasting point pp ? client
play-out point
8
Solutions for Providing VCR-like Functions

• Some proposed solutions
• Channel merging
• Pre-fetching / Buffering
• Staggered broadcasting
• Quality degradation
• Precision reduction
• Movie preview

9
Channel Merging

• Goal to enjoy server bandwidth save by multicast
  by merging streams together

bp
client buffer
pp
forward VCR operation
bp
client buffer
pp
channel merging
bp
client buffer
pp
Examples patching, piggybacking etc.
10
Patching
bp
tL
movie
pp
bp
client buffer
pp
multicast channel
transient period (T tL)
contingency channel
time
bp
client buffer
pp
K. A. Hua, Y. Cai and S. Sheu, "Patching A
Multicast Technique for True Video-on-Demand
Services," Proc. 6th International Conference on
Multimedia , Sept 1998 Page(s) 191-200.
11
Patching

• Gain
• Further reduction of response time
• Multicast efficiency improvement
• Simulation results
• No. of server channels 95 vs 100 for general
  multicast systems (arrival rate 0.1 / s)
• Blocking probability 8 vs 13 for general
  multicast systems (arrival rate 0.1 / s)
• Trade-off
• Access bandwidth 2X movie rate
• Client buffer (max size Tr of movie data, for
  staggered broadcasting)

Ho Kyun Park, Hwang Bin Ryou, Multicast
Delivery for Interactive Video-on-Demand
Service, Proc. 12th International Conference on
Information Networking, 1998 Page(s) 46 -50.
12
Patching

• Variant
• Playback rate of contingency channel SRMDRU
  (Single Rate Multicast Double Rate Unicast) Bu
  2
• Transient duration halved (T tL/2)
• Requires access bandwidth of 3X movie rate
• Desirable application
• Small phase offset (tL) between client play-out
  point and multicasting point ? short patching
  duration

Poon, W.F., Lo, K.T., Feng, J., Design and
analysis of multicast delivery to provide VCR
functionality in video-on-demand systems, 2nd
International Conference on ATM, 1999, Page(s)
132 -139.7
13
Piggybacking

• Algorithm
• To merge two streams by altering display rate
• Gain
• Enables stream merging without the use of
  contingency channels and client buffer
• Trade-off
• Long merging duration (T 10tL for r 5)

tL
pp1
stream 1 at (1-r) movie rate
transient period (T tL / 2r)
stream 2 at (1r) movie rate
pp2
pp1
merged and share muticast
r ? display rate alternation ratio
pp2
L. Golubchik, J. C. S. Lui, and R. R. Muntz,
"Adaptive Piggybacking A Novel Technique for
Data Sharing in Video-on-Demand Storage Servers,"
ACM Multimedia Systems, vol.4(30), 1996 Page(s)
14-55.
14
Pre-fetching / Buffering

• Algorithm
• Always keeps the pp in the middle of buffered
  video

bp2
bp1
bp4
bp3
bp2
bp1
bp4
bp3
pp
pp
Zongming Fei Kamel, I. Mukherjee, S. Ammar,
M.H., Providing Interactive Functions for
Staggered Multicast Near Video-on-Demand
Systems, IEEE International Conference on
Multimedia Computing and Systems, Volume 2 ,
1999, Page(s) 949 -953.
15
Pre-fetching / Buffering

• Gain
• Higher probability of buffer hit ? VCR operation
  completion
• Simulation result
• 92 vs 63 for conventional buffer schemes
• Trade-off
• Extra client access bandwidth requirement (max.
  3X)
• Large client side buffer size (3 Tr of movie data
  Tr L / Ns)
• Desirable application
• Used together with periodic broadcasting

16
Staggered Broadcasting

• Algorithm
• Gain
• An upper bound on buffer size for merging (Tr)
• An upper bound on resumption waiting time (Tr,
  mean Tr / 2)
• Trade-off
• Wastage of server bandwidth in case of batch size
  0 (no req. arrival in a whole time slot)
• Desirable application
• Medium to high arrival rate (arrival rate gt 1/Tr)

17
Quality Degradation

• Algorithm
• Quality of movie playback, such as frame rate or
  resolution is lowered during transient period
• Gain
• Lower client access bandwidth / server bandwidth
  requirement
• Trade-off
• Lower movie quality
• Sophisticated hardware or pre-coding may be
  required for production of the altered video
  stream
• Desirable application
• Efficient transcoding techniques available

18
Precision Reduction

• Algorithm
• Restricts play-out point jumps to video
  broadcasting points
• Gain
• No additional contingency requirement
• Zero buffer requirement
• Trade-off
• Resumption points after VCR-like operations are
  in the increment of t (e.g. 5 mins.)
• Desirable application

bp2
bp1
bp4
bp3
bp2
bp1
bp4
bp3
movie
pp
pp
pp
Almeroth, K.C. Ammar, M.H., On the Performance
of a Multicast Delivery Video-on-Demand Service
with Discontinuous VCR Actions, IEEE
International Conference on Communications,
Seattle, 'Gateway to Globalization', Volume 3 ,
1995, Page(s) 1631 -1635.
19
Previewing Movie

• Algorithm
• Shows preview during admission / VCR resumption
  downloading / waiting period
• Gain
• Creates an illusion which shortens clients
  perceived waiting time
• Let clients confirm the right movie selection /
  seek point
• Trade-off
• Zero or even slightly negative bandwidth save,
  depending on reusability of the preview data
• Desirable application
• Client is content with movie preview

Wallapak Tavanapong , Kien A. Hua, James Z.
Wang, A Framework for Supporting Previewing and
VCR Operations in a Low Bandwidth Environment, ,
Proceedings of the fifth ACM international
conference on Multimedia, November 1997
20
Algorithm Design

• Main VoD system design considerations
• Quality of service
• Access latency
• VCR-like function resumption destination shift
• Movie playback quality
• FF / FB speed-up factor
• STB design
• Client access bandwidth
• Client buffer size
• Channel retrieval policy
• Server design
• Channel scheduling complexity
• Server bandwidth

21
Algorithm Design

• Trade-offs of different techniques
• Compared to an on-demand batching multicast VoD
  system.

22
Algorithm Design

• Different approaches work well under different
  system parameters
• Staggered broadcasting is suitable for moderate
  to high request arrival rate
• Real-time video transcoding necessary for
  Quality Degradation approach is possible only
  on high-end servers

23
Algorithm Design

• We consider an adaptive hybrid approach
• Adaptive ? different combinations of algorithms
  utilized reacting to change in admission / VCR
  resumption request arrival pattern
• Hybrid ? different combinations of algorithms
  utilized depending on the initial system
  parameter restrictions

24
Current Work

• Current proposed system
• Based on SS-VoD, uses a combination of 3
  approaches
• Staggered broadcasting
• Batched patching
• Precision reduction

C.H. Lee, Y. B. Lee, Design, Performance
Analysis and Implementation of a Super-Scalar
Video-on-Demand System
25
Current Work Staggered Broadcasting

• SS-VoD architecture
• Staggered broadcasting by static multicast
  channels
• Dynamic multicast channels for admission patching
  and VCR-like functions resumptions
• Problem
• Even a small (0.7 time per client) Prob(VCR-op)
  saturates the whole system (1 movie, L 9500,
  25/25 channel allocation, arrival rate 0.1 /s)
  (admission wait 6.58 ? 143.01)

dynamic multicast channels
26
Current Work Batched Patching

• Motivation
• Attempt to minimize server loading due to VCR
  resumptions by multicasting
• Algorithm
• Admission / merging requests of the same target
  point are served by one single dynamic multicast
  channel
• Result
• Able to reduce channel requirement for dynamic
  admission
• Problem
• VCR resumption request target points are sparsely
  distributed in time domain, and it is impossible
  to form a batch

merging request 1
batched merging request 1
merging request 2
27
Current Work Precision Reduction

• Motivation
• Try to maximize VCR resumption batch size by
  reduction of seeking accuracy
• Algorithm 1
• All VCR resumption requests fall within a window
  of pp /- tw are served as a batch

pp1
tw
tw
tw
tw
batching
pp2
pp1 pp2
28
Current Work Precision Reduction

• Algorithm 2
• Restrict clients to seek to predefined points
  only (chapter-based seeking)
• Result
• Further reduce bandwidth requirement by batching

pre-defined chapter seek points
movie
L
29
Future Work

• Method to further increase batch size
• Investigation of applicability of transition
  patching (recursive patching) on current system

Ying Cai , Kien A. Hua, An efficient
bandwidth-sharing technique for true video on
demand systems, Proceedings of the seventh ACM
international conference on Multimedia, 1999 ,
Orlando, Florida, United States
30
Q A

• Thank you