Issues in designing a distributed hierarchical storage system for continuous media service.

摘要

Recent advances in computing and communication technologies have made it technically feasible and economically viable to provide on-line access to a variety of information services over high speed networks. Particularly, convergence of various technological factors, namely in network access and in video coding and transmission, have recently brought a rapid growth of interest in on-line access to multimedia services. Large-scale provision of such service is attractive to end users as well as service providers. However, excessive bandwidth and space requirements of continuous media data have been a serious impediment to the wide spread usage of on-line continuous media services, especially in the commercial entertainment market. In this dissertation, we propose distributed hierarchical storage architecture as a promising solution to cost-effective service provisioning. The advantage of adopting a hierarchical storage architecture is its ability to assign the appropriate storage hierarchy to each file based on its the access frequency. However, success in using a storage hierarchy relies upon selecting appropriate operational parameters with a given user access profile. Our hierarchical storage system consists of primary storage, secondary storage, and tertiary storage. This dissertation analyzes the performance of the hierarchical storage system under various combinations of system attributes and develops a technique to find the minimum amount of resources in each level of hierarchies while satisfying certain operational constraints. The second part of the dissertation focuses on distributed service provisioning. Techniques are developed to exploit the geographical locality of reference and temporal locality of reference by introducing intermediate storages. Due to non-trivial network resource requirements, it is beneficial to put the data closer to the end user. To precisely compute the trade-offs between network resource consumption and storage resource cost, we develop a cost model which quantifies the aggregate resource consumption. On the basis of the cost model, an algorithm which finds the efficient way of servicing a set of requests is developed. The observations and findings from the mathematical models are validated with the simulation studies.