Saleable Techniques for Video on Demand

1
Saleable Techniques forVideo on Demand

• Kien A. Hua
• School of EE Computer Science
• University of Central Florida
• Orlando, FL 32816-2362
• U.S.A

2
Server Channels

• The unit of server capacity required to deliver a
  video stream is referred to as a channel.
• These channels are dispatched to deliver various
  video streams at different times.

3
Using Dedicated Streams
Video Server
Dedicated stream
Client
Expensive
Client
Client
Not scalable
Client
4
A Solution

• Using broadcast to share channels among users
• Broadcast is essentially free for a large user
  community

5
Traditional Multicast
Video Server
Client
Client
Client
Client
6
Conflicting Goals in Video
Multicast ?

• Low Latency requests must be served as soon as
  possible

• High Efficiency each multicast should wait to
  serve a larger number of clients

Can we achieve both ?
7
Broadcast for VOD

• Requirement on server bandwidth is independent
  of the number of users the system is designed to
  support.

Less expensive more scalable !!
Broadcast cannot deliver videos on demand ?
?
?
True False
8
Simple Periodic BroadcastStaggered Broadcasting

• A new stream is started every interval for each
  video.
• The worst service latency is the broadcast period.

9
Limitation of Simple Periodic Broadcast

• Access latency can be improved only linearly with
  increases to the server bandwidth
• Double the number of channels to reduce the
  service latency in half

Can we do better ?
10
Skyscraper Broadcasting

• Each video is fragmented into K segments, each
  repeatedly broadcast on a dedicated channel at
  the playback rate.
• The sizes of the K segments have the following
  pattern
• 1, 2, 2, 5, 5, 12, 12, 25, 25, , W, W, , W

Odd group
Even group
Odd group
Even group
Size of larger segments are constrained to W
(width of the skyscraper)
11
Generating Function

• The broadcast series is generated using the
  following recursive function

1 If n 1,
2 If n 2 or 3,
2 f(n - 1)1 If n mod 4 0.
f(n-1) If n mod 4 1,
2 f(n - 1) 2 If n mod 4 2,
f(n 1) If n mod 4 3
f(n)
12
Skyscraper Broadcasting Playback Procedure

• The Odd Loader and the Even Loader download the
  odd groups and the even groups, respectively.
• The W-segments are downloaded sequentially using
  only one loader.
• As the loaders fill the buffer, the Video Player
  consumes the data in the buffer.

13
SB Multiuser Example
1
2
3
4
5
6
7
8
Playback schedule

• Each segment is available before it is needed for
  the playback

14
Advantages of Skyscraper Broadcasting

• Unlimited scalability
• Service latency can be reduced exponentially with
  increases in server bandwidth
• Since W-segments are downloaded sequentially,
  buffer requirement is minimal.

Skyscraper
15
Client Centric Approach (CCA)

• Segments are grouped according to the number of
  channels the clients can download simultaneously
• Inside each group, each segment is twice the size
  of the last segment
• The first segment of any group is the same size
  as the last segment of the previous group

C 3 (Clients have three loaders)
16
CCA Broadcasting

• Server broadcasts each segment at playback rate
• Clients use c loaders
• Each loader downloads its streams sequentially,
• e.g., i th loader is responsible for segments i,
  ic, i2c, i3c,
• Equal-size W-segments are downloaded sequentially
  using one loader

C 3 (Clients have three loaders)
17
A CCA Example
18
Advantages of CCA

• It has the advantages of Skyscraper Broadcasting.
• It can leverage client bandwidth to improve
  performance.

19
UCF Video Jukebox
20
Cautious Harmonic Broadcasting(Segmentation
Design)

• A video is partitioned into n equally-sized
  segments
• The first channel repeatedly broadcasts the first
  segment S1 at the playback rate
• The second channel alternately broadcasts S2 and
  S3 repeatedly at half the playback rate
• Each of the remaining segment Si is repeatedly
  broadcast on its dedicated channel at 1/(i1) the
  playback rate

21
Cautious Harmonic Broadcasting(Download
Playback Strategy)

• The client receives data from all streams for
  the video simultaneously.
• The client starts the playback as soon as it can
  download the first stream.

22
Cautious Harmonic Broadcasting

• Advantage Better than Skyscraper Broadcasting in
  terms of server bandwidth requirement.
• Disadvantage Requires many times more client
  bandwidth then Skyscraper Broadcasting does.
• Implementation Issues
• The client must receive data from many channels
  simultaneously (e.g., 240 channels are required
  for a 2-hour video if the desired latency is 30
  seconds).
• A storage subsystem capable of moving the read
  heads fast enough to multiplex among so many
  concurrent streams would be very expensive, if
  possible

23
Pagoda BroadcastingData Fragmentation

• Each video is divided into equally-sized segments
• Using the following series to determine the
  number of segments for each channel
• 1, 3, 5, 15, 25, 75, 125,
• Segments appearing on a channel do not have to be
  consecutive

3rd segment
24
Pagoda BroadcastingDownload and Playback Strategy

• Each channel broadcasts data at the playback rate
• The client receives data from all channels
  simultaneously.
• It starts the playback as soon as it can download
  the first segment.

25
Pagoda BroadcastingAdvantage Disadvantage

• Advantage Server bandwidth requirement is
  lower compared to Skyscraper Broadcasting
• Disadvantage Client bandwidth requirement
  is many times higher than Skyscraper Broadcasting
• Achieving a maximum delay of 138 seconds for a
  2-hour video requires each client to have a
  bandwidth five times the playback rate, e.g.,
  approximately 20 Mbps for MPEG-2
• System cost is significantly more expensive

26
New Pagoda Broadcasting

• New Pagoda Broadcasting improves on the original
  Pagoda Broadcasting.
• Client bandwidth requirement remains very high
• Example Achieving a maximum delay of 110
  seconds for a 2-hour video requires each client
  to have a bandwidth five times the playback rate.
• Approximately 20 Mbps for MPEG-2
• System cost is very expensive

27
Total System Cost

• Service latency can be improved by increasing
  bandwidth in two ways
• Increasing client bandwidth, e.g., Harmonic
  Broadcasting, Pagoda Broadcasting, etc.
• This approach is expensive
• Increasing server bandwidth, e.g., Skyscraper
  Broadcasting, CCA, etc.
• This approach is less expensive

28
Support Client Heterogeneity

• Using multi-resolution encoding
• Bandwidth Adaptor
• HeRO Broadcasting

29
Multi-resolution Encoding

• Encode the video data as a series of layers
• A user can individually mould its service to fit
  its capacity
• A user keeps adding layers until it is congested,
  then drops the higher layer

Drawback Compromise the display quality
30
Bandwidth Adaptors
Advantage All clients enjoy the same quality
display
31
Requirements for an Adaptor

• An adaptor dynamically transforms a given
  periodic broadcast into another less demanding
  one
• The segmentation scheme must allow easy
  transformation of a broadcast into another
• CCA segmentation technique has this property

32
Two Segmentation Examples
33
Adaptation (1)
Adaptor downloads from all broadcast channels
simultaneously
34
Adaptation (2)

• Each sender routine retrieves data chunks from
  buffer, and broadcast them to the downstream
• For each chunk, the sender routine calls
  deleteChunk to decide if the chunk can be deleted
  from the buffer

35
Buffer Management

• insertChunk implements an As Late As Possible
  policy, i.e.,
• If another occurrence of this chunk will be
  available from the server before it is needed,
  then ignore this one, else buffer it.
• deleteChunk implements an As soon As Possible
  policy, i.e.,
• Determine the next time when the chunk will need
  to be broadcast to the downstream.
• If this moment comes before the availability of
  the chunk at the server, then keep it in storage,
  else delete it.

36
The Adaptor Buffer

• Computation is not intensive.
• It is only performed for the first chunk of the
  segment, i.e.,
• if this initial chunk is marked for caching, so
  will be the rest of the segment.
• Same thing goes for deletion.

37
The start-up delay

• The start-up delay is the broadcast period of the
  first segment on the server

38
HeRO Heterogeneous Receiver-Oriented
Broadcasting

• Allows receivers of various communication
  capabilities to share the same periodic broadcast
• All receivers enjoy the same video quality
• Bandwidth adaptors are not used

39
HeRO Data Segmentation

• The size of the i th segment is 2i-1 times the
  size of the first segment

40
HeRO Download Strategy

• The number of channels needed depends on the time
  slot of the arrival of the service request
• Loader i downloads segments i, iC, i2C, i3C,
  etc. sequentially, where C is the number of
  loaders available.

Global Period
41
HeRO Regular Channels

• The first user can download from six channels
  simultaneously

Request 1
42
HeRO Regular Channels

• The second user can download from two channels
  simultaneously

Request 2
43
Worst-Case for Clients with 2 loaders

• Worst-case latency is 11 time units
• The worst-cases appear because the broadcast
  periods coincide at the end of the global period

Request 2
11 time units

• Coincidence of the
• broadcast periods
• require more
• loaders

44
Worst-Case for Clients with 3 loaders

• Worst-case latency is 5 time units
• The worst-cases appear because the broadcast
  periods coincide at the end of the global period

Request
5 time units
Coincidence of the broadcast periods
45
Observations of Worst-Cases

• For a client with a given bandwidth, the time
  slots it can start the video are not uniformly
  distributed over the global period.
• The non-uniformity varies over the global period
  depending on the degree of coincidence among the
  broadcast periods of various segments.
• The worst non-uniformity occurs at the end of
  each global period when the broadcast periods of
  all segments coincide.
• The non-uniformity causes long service delays for
  clients with less bandwidth.
• We need to minimize this coincidence to improve
  the worst case.

46
Adding one more channel

• We broadcast the last segment on one more
  channel, but with a time shift half its size.
• We now offer more possibilities to download the
  last segment and above all, we eliminate the
  coincidence of all segments (i.e., no longer
  requiring 6 loaders).

Regular Group
Shifted Channel
47
HeRO

• To reduce service latency for less capable
  clients, broadcast the longest segments on a
  second channel with a phase offset half their
  size.

Shifted Channels
48
HeRO Experimental Results

• Under a homogeneous environment, HeRO is
• competitive in service latencies compared to
  other protocols
• the most efficient protocol to save client buffer
  space
• HeRO is the first periodic broadcast technique
  designed to address the heterogeneity in receiver
  bandwidth
• Less capable clients enjoy the same playback
  quality

49
Hybrid Approach

• Periodic broadcast is better for very popular
  videos
• Batching (scheduled multicast) is more
  appropriate for less popular videos
• A hybrid of these two approaches
• offers the best performance

50
Adaptive Hybrid Approach (AHA)

• Popularity of each video is assessed periodically
  based on the distribution of recent requests.
• Popular videos are served using Skyscraper
  Broadcasting
• Less popular videos are served using batching
• Number of channels used for periodic broadcast
  depends on the current mix of popular videos
• Remaining channels are allocated to batching

51
AHA Determine Popularity

• A video Vi is popular if the following two
  conditions are true
• Test 1 verifies that Vi is relatively popular to
  deserve the periodic broadcast channels
• Test 2 ensures that Vi is indeed popular (i.e.,
  small inter-arrival rate)

Number of channels required for periodic broadcast
Worst delay for Staggered Broadcasting
52
Batching (Scheduled Multicast)
A channel is available
Channel Pool
Batching Scheduler
Multicast
53
AHA Schedulers

• Popularity Evaluator periodically assigns each
  video to either Batching Scheduler or Broadcast
  Scheduler
• When a request arrives for a popular video,
  Broadcast Scheduler informs the client which
  channels to download the video
• For less popular videos, Batching Scheduler
  assigns the next available channel to multicast
  the video with largest aggregated waiting time

54
We can use periodic broadcast for VOD
Can we use scheduled multicast for VOD ?
55
Patching First Client
Video
Regular Multicast
A
56
Patching Second Client
Skew point
Video
t
Patching Stream
Regular Multicast
A
57
Patching Clients Arrive at Different Times
Video
2t
Regular Multicast
Skew point is absorbed by client buffer
Buffer
Video Player
A
B
58
Optimal Patching Window
time
patching window W
patching window W
r
p
p
p
r
p
p
A
B
C
D
E
F
G
Multicast group
Multicast group
What is the optimal patching window ?
59
Simple Patching

• Patching is used if the client has enough buffer
  space to absorb the skew.
• Too greedy. May result in many long patching
  streams

60
Optimal Patching Window

• Compute D, the mean amount of data transmitted
  for each multicast group (a function of the
  patching period W )
• Determine ? , the average time duration of a
  multicast group (a function of W )
• Server bandwidth requirement is D/? which is a
  function of the patching period
• Finding the patching period W that minimizes the
  bandwidth requirement (D/?)

61
Candidates for Optimal Patching Window
62
Limitation of Patching

• Performance of Patching is limited by server
  bandwidth.
• Can we scale the application beyond the physical
  limitation of the server ?

63
Range Multicast
9
10
Video Server
5
6
7
8
4
RM Router 1
RM Router 2
1
2
3
64
Range Multicast
65
Multicast Range

• All members of a conventional multicast share the
  same play point at all time
• They must wait until the multicast time
• Members of a range multicast can have a range of
  different play points
• They can join a range multicast at their leisure
  without waiting

Multicast Range at time 11 0, 11
66
Prototype RM Routers
67
Support VCR-like Operations in a Broadcast
Environment ?
68
VCR-Like Interactivity

• Continuous Interactive functions
• Fast forward
• Fast rewind
• Pause
• Discontinuous Interactive functions
• Jump forward
• Jump backward
• Useful for many VoD applications

69
VCR Interaction Using Client Buffer
Video stream
Video stream
Video stream
Video stream
Video stream
70
Interaction Using Batching

• Requests arriving during a time slot form a
  multicast group
• Jump operations can be realized by switching to
  an appropriate multicast group
• Use an emergency stream if a destination
  multicast group does not exist

Jump
Jump
71
Continuous Interactivity under Batching

• Pause
• Stop the display
• Return to normal play as in Jump
• Fast Forward
• Fast forward the video frames in the buffer
• When the buffer is exhausted, return to normal
  play as in Jump
• Fast Rewind
• Same as in fast forward, but in reverse direction

72
Split and Merger (SAM) Protocol

• Uses 2 types of streams, S streams for normal
  multicast and I streams for interactivity.
• When a user initiates an interactive operation
• Use an I channel to interact with the video
• When done, use the I channel as a patching stream
  to join an existing multicast
• Return the I channel

Advantage Unrestricted fast forward and rewind
Disadvantage I stream, used by one user, is
expensive
73
Resuming Normal Play in SAM

• Use the I stream to download segments 6 and 7,
  and render them onto the screen
• At the same time, join the target multicast and
  cache the data, starting from segment 8, in a
  local buffer

The user just finished the interactivity
74
Interaction with Broadcast Video

• The interactive techniques developed for Batching
  can also be used for Staggered Broadcast
• However, Staggered Broadcast does not perform
  well

75
Client Centric Approach (CCA)

• Server broadcasts each segment at the playback
  rate
• Clients use c loaders
• Each loader downloads its streams sequentially,
• e.g., i th loader is responsible for segments i,
  ic, i2c, i3c,
• Equal-size W-segments are downloaded sequentially
  using one loader

C 3 (Clients have three loaders)
76
CCA is Good for Interactivity

• Segments in the same group are downloaded at the
  same time
• Facilitate fast forward
• The last segment of a group is of the same size
  as the first segment of the next group
• Ensure smooth continuous playback after
  interactivity

77
Broadcast-based Interactive Technique (BIT)
Compression ratio is 4
78
BIT

• Two Buffers
• Normal Buffer
• Interactive Buffer
• When Interactive Buffer is exhausted, client must
  resume normal play

79
BIT Loaders Receiving Data

• c2 loaders, where c is the group size
• c regular loaders download the regular segments
  as in CCA
• The 2 interactive loaders function as follows
• If the current playback segment is in the first
  half of its group, the two interactive loaders
  are allocated to the previous and current
  interactive segments
• Otherwise, they are allocated to the current and
  next interactive segments

Keeping play point at middle of interactive buffer
80
BIT Resume-Play Operation
Tune to these three segments simultaneouslyActua
l destination point is chosen from among frames
at broadcast point to ensure continuous playback
81
BIT Resume-Play Operation(All Scenarios)
Tune to these three segments simultaneouslyActua
l destination point is chosen from among frames
at broadcast point to ensure continuous playback
82
BIT - User Behavior Model

• mx duration of action x
• Px probability to issue action x
• Pi probabilty to issue interaction
• mi duration of the interaction
• mff mfr mpause mjf mjb,
• Ppause Pff Pfb Pjf Pjb Pi/5.
• dr mi/mp interaction ratio.

83
Two Performance Metrics

• Percentage of unsuccessful action
• Interaction fails if the buffer fails to
  accommodate the operation
• e.g., a long-duration fast forward pushes the
  play point off the Interactive Buffer
• Average Percentage of Completion
• Measure the degree of incompleteness
• e.g., if a 20-second fast forward is forced to
  resume normal play after 15 seconds, the
  Percentage of Completion is 15/20, or 75.

84
BIT - Simulation Results
85
P2P Live Broadcast
86
Live Broadcast

• Video on Demand Each user plays the video from
  the beginning
• Live Video Broadcast A user joins an on-going
  live broadcast at the current play point of the
  video.
• ZIGZAG is a peer-to-peer technique for live
  broadcast

87
Chaining First P2P Streaming
P2P Streaming is another way to scale beyond the
bandwidth limitation of the server
88
Requirements in P2P Live Broadcast
89
Liveness Load Balancing
Server
new
90
Robustness
The server is already busy
Server
Too many reconnections!
91
Control Overhead

• Peers periodically exchange information to
  maintain its position and connections.
• The overhead of this task should be small and
  independent of the number of peers

92
ZIGZAG Live Broadcast

• A peer, at its highest level, connects to peers
  from a different cluster in the next level
• Peers that are not leaders get data from a leader
  of a different cluster

93
ZIGZAG - Advantages

• Short broadcast tree improves liveness
• Small fanout keeps reconnecting cost low
• Hierarchy helps reduce control message overhead

94
Pull-based P2P Streaming

• Join the network as a new peer
• Obtain information of desired data through a
  gossip-based mechanism
• Pull data from different peers

95
Pull vs Push

• Push
• Peers at the edges of the multicast structures do
  not contribute resources
• Pull
• A lot of redundant data in the network
• Searching for data (gossiping or using DHT)
  incurs overhead

96
Many P2P Streaming Systems on Internet

• PPLive
• PopCast
• PPStream
• Coolstreaming
• TV Ants
• QQLive

97
2-Phase Service Model(2PSM)Browsing Video in
a Low Bandwidth Environment
98
Search Model

1. Use similarity matching or keyword search to look
   for the candidate videos.
2. Preview some of the candidates to identify the
   desired video.
3. Apply VCR-style functions to search for the video
   segments.

99
Conventional Streaming
Advantage
Reduce wait time
Disadvantage
Not suitable for searching video
100
Search Techniques

• Use extra preview files to support the preview
  function
• Requires more storage space
• Downloading the preview file adds delay
• Use separate fast-forward and fast-reverse files
  to provide VCR-style operations
• Requires more storage space
• Server can become a bottleneck

101
VCR-like Functionality is Expensive

• Supporting VCR-like operations
• demands substantial network bandwidth, and
• requires significant server resources
• Can we avoid these costs ?

102
2PSM Preview Phase
103
2PSM Playback Phase
t
104
Advantages
1. It requires no extra files to provide the
preview feature.
2. Downloading the preview frames is free.
3. It requires no extra files to support the VCR
functionality.
4. Server is not involved in the VCR-style
interaction.
105
UCF Video Browser
106
iSEE - Intelligent Sensors Exploration
Environment