Digital Library Accessing Techniques

1
Digital Library Accessing Techniques
Supervisors Prof. Michael Lyu, Prof. Irwin
King Presenter Lee Chi Wai, Anson Date
7-12-1999
2
Outline of the presentation

• Introduction
• Bandwidth smoothing algorithms
• Mobile agent communication model
• Audio server/client delivery system
• Conclusion
• Question and answer

3
Introduction

• Live video delivery techniques
• Challenges decisions must be made in real-time
  and delay must be minimized
• Solution weaker guarantees of service or
  degradation in quality of service
• Stored video delivery techniques
• used in digital libraries and video-on-demand
  systems

4
Introduction

• Stored video delivery techniques(cont.)
• the full video is know in advance
• use pre-fetching techniques to control the
  picture quality
• What is pre-fetching techniques?
• create a pre-fetch buffer in client side
• fill the pre-fetch buffer with more bytes in
  advance of each burst

5
Introduction

• How to fill the pre-fetch buffer?
• Server uses some bandwidth smoothing algorithms
• Critical bandwidth allocation (CBA)
• introduced by W. Feng and S. Sechrest in 1995
• Minimum changes bandwidth allocation (MCBA)
• introduced by W. Feng and S. Sechrest in 1995

6
Introduction

• Bandwidth smoothing algorithms (cont.)
• Minimum variability bandwidth allocation
• introduced by J. Salehi, Z. Zhang, J.Kurose, and
  D. Towsley in 1996
• Piecewise Constant Rate Transmission and
  Transport Algorithm
• introduced by J. Mcmanus and K.Ross in 1996

7
Introduction

• Problem of the client/server communication model
  
• static and inflexible
• interaction between clients and servers requires
  permanent network connections
• cant support mobile computers

8
Introduction

• Solution
• use Mobile Agent communication model
• support asynchronous interaction of searching for
  information
• support for mobile computers or lightweight
  devices

9
Bandwidth smoothing Algorithm

• Aim reduce the burstiness of the network
  traffic by pre-fetching video data to the client
  side at a series of fixed rates in advance of
  each burst (transmission plan)
• Different bandwidth smoothing algorithms use
  various strategy to generate their transmission
  plan

10
Transmission plan

• Suppose a compressed video stream consists of n
  frames, where frame I requires fi bytes of
  storage
• The server must always transmit quickly enough to
  avoid pre-fetch buffer underflow, where

11
Transmission plan

• The server must not transmit more data than
• It prevent the overflow of the pre-fetch buffer
  (of size b)

12
Transmission plan

• Hence, any valid transmission plan should follow
  the following constraint
• where ci is the transmission rate during frame
  slot i of the smoothed video stream

13
Transmission plan

• Creating a transmission plan involves generating
  m consecutive runs each with a constant bandwidth
  allocation rj and duration tj

14
Critical bandwidth allocation

• For the need of decrease bandwidth, the CBA
  algorithm starts a rate decrease at the leftmost
  point of the frontier

15
Critical bandwidth allocation

• For the need of increase bandwidth, the CBA
  performs a search along the frontier to locate
  the starting point that allows the next
  trajectory to extend as far as possible

16
Critical bandwidth allocation

• smallest possible peak bandwidth requirement
• minimum number of bandwidth increases
• does not necessary have the minimum number of
  bandwidth decreases, since the algorithm selects
  the leftmost starting point

17
Minimum changes bandwidth allocation

• Extends the CBA scheme to perform the linear
  search operation on all rate changes
• smallest possible number of rate changes
• minimum peak bandwidth
• requirement

18
Minimum variability bandwidth allocation

• Reduce the variability in the rate requirements
  across the lifetime of the transmission plan
• Initiates bandwidth changes at the leftmost point
  along the frontier, for both rate increases and
  rate decreases

19
Run boundary example

• Selection of starting point of different
  bandwidth smoothing algorithm

20
Piecewise Constant Rate Transmission and
Transport Algorithm

• Aim Generate a single run for each intervals by
  connecting the interaction points on the Funder
  curve

21
Piecewise Constant Rate Transmission and
Transport Algorithm

• Slope of the dotted line corresponds to the rate
  rj in the resulting transmission plan
• vertically offsets the plan until all runs lie
  above Funder curve
• introducing an initial playback delay in the
  pre-fetch buffer

22
Mobile agent

• Definition of mobile agent
• a program that is able to migrate from node to
  node on a network under its own control for the
  purpose of completing a task specified by a user
• The agent chooses when and to where it will
  migrate and may interrupt its own execution and
  continue anywhere on the network

23
Reasons for using Mobile Agent

• Asynchronous interaction by sending one active
  message to the service site
• Reduction of network traffic

24
Reasons for using Mobile Agent

• Support for mobile and lightweight devices

25
Audio Server/Client delivery system
26
Audio Server/Client delivery system

• The system is implemented under Linux
• Use MPEG 1 layer 3 compression technique to
  compress the audio file
• mp3 files are stored in the same machine where
  the audio server is running

27
Transmission plan

• mp3 bit rate 128kbps
• try to approximate the bit-rate by introducing
  sleep intervals between successive sending
  attempts
• Maximum Transfer Unit(MTU) 512 bytes
• Sleep interval MTU/(mp3 bit rate)
• 32000?sec

28
Transmission plan
While (server buffer not empty) sleep interval
0.03125sec Record the start time Transmit
audio packet Record the stop time Sleep for
sec
29
Buffer Management in Client Side

• Need to buffer audio data at the beginning of
  playback process since some packets may arrive
  too late
• Or some packets may arrive in burst
• Use Leaky bucket Model in the client side buffer
  to produce a constant output bit rate (128kbps)
  to the decoder

30
Conclusion

• We have presented four bandwidth smoothing
  algorithms
• We have discussed the possibility of substituting
  client/server model with mobile agent
  communication model
• The audio server/client delivery system has been
  presented

31
Future Work

• Implement video server/client delivery system
• Use mobile agent in the multimedia applications
  to delivery audio or video data

32
Question and Answer