Physical storage model for interactive multimedia information systems
Abstract
A framework for the storage layout design environment for interactive multimedia information systems is introduced in this thesis. We propose a two-level conceptual model that can easily capture the behavior of user interactions such as cyclic browsing, access statistics, and temporal relations among multimedia objects. A large data placement problem can be divided into a number of small data placement problems by weighted graph decomposition. The Fiduccia-Mattheyses partitioning algorithm is recursively applied for this purpose. Once the graph is fully partitioned, the objects in the same subgraph are assigned to the same disk. The expected access patterns of browsing are modeled as a time-homogeneous ergodic Markov Chain, from which the stationary probability for each node of the browsing graph can be found. Based on these probabilities, we define an expected access cost and propose various clustering and storage layout algorithms. The problem of clustering data for cyclic browsing is posed as an optimization problem, which is computationally equivalent to the linear placement problem--an NP-complete problem. Our simulation study has revealed that the Maximum-cost Chain Connection (MCC) algorithm can generate a layout close to the optimal placement, especially for browsing graphs with low connectivity. In the detailed data placement, storage layout for sequential (linear) multimedia presentation must satisfy the synchronization constraints among multiple concurrent data streams to allow a smooth flow of data and preservation of originality. To achieve that, we propose three detailed data placement algorithms and retrieval/decoding schedule which allow us not only to meet the synchronization constraints of the multimedia application but also maximize the data transfer bandwidth of the storage device by reducing the effects of disk seek time and rotational latency. Conventional file servers and associated storage systems can not support multimedia applications such as video services on local area networks. Two data placement approaches based on the interleaving and the non-interleaving strategies are developed to guarantee real-time playback for multimedia applications stored disk arrays.
Degree
Ph.D.
Advisors
Kashyap, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.