MultiQuality Data Replication in Multimedia Databases

1
Multi-Quality Data Replication in Multimedia
Databases

• Yi-Cheng Tu, Student Member, IEEE, Jingfeng Yan,
  Student Member, IEEE,
• Gang Shen and
• Sunil Prabhakar, Senior Member, IEEE
• Presenter
• Kishore Nallagatla
• CSU ID 2460963

2
Introduction to Multimedia Databases

• Quality is the most essential property for a
  Multimedia Database
• Multimedia Databases can have a wide range of
  quality parameters
• Quality aware multimedia systems allow users to
  specify the quality of media to be delivered

3
Approaches to Satisfy User Quality Specifications

• Dynamic Adaptation
  Store only the highest resolution copy and
  convert it to the quality format requested
• Static Adaptation Precompute each different
  quality that can be requested and store them on
  disk

4
Advantages Disadvantages

• In Dynamic adaptation, a very high CPU overhead
  is required for transcoding from one quality to
  another
• In Static adaptation, a high disk space is
  required to store replicas of different qualities
• We trade disk space for runtime CPU cycles, which
  is a cost-effective trade-off since disks are
  relatively cheap

5
Assumptions in Static adaptation

• User requests concentrate on small number of
  quality profiles
• There is always enough storage space

6
The Problem

• The choice of which quality copies to store
  becomes important and is the focus of this paper
• This is called the Replication problem

7
Traditional Data Replication

• The copies of data are placed in various nodes in
  a distributed environment

8
Quality-Aware Replication

• The data is placed in a metric space of quality
  values termed as quality space

9

• In this paper, strategies are presented to
  choose the quality of replicas under two
  different user requirements
• Hard Quality Model User receives the exact
  quality required
• Soft Quality Model Users are willing to
  negotiate quality they will receive

10
Notations Used In The Paper
11
Strategy for Hard-Quality Systems

• The unavailability of a requested quality results
  in rejection of the request
• Hence the reject probability p will be a metric
  for performance evaluation

12
Strategies for Soft-quality Systems

• Utility Functions
• Data Replication as an optimization

13

• The Greedy algorithm

14

• The Iterative Greedy Algorithm

15
Conclusion

• In this paper, the problem of selecting quality
  specific replicas of media data has been studied
• This problem is generally overlooked in
  multimedia database research due to the
  assumption that storage space is abundant
• In this paper is has been demonstrated that this
  is not the case if the system is to adapt to user
  quality requirements