Hierarchical data placement for navigational multimedia applications
Date
2003Author
Βακάλη, Αθηνά
Τερζή, Εβημαρία
Bertino, Elisa
Elmagarmid, Ahmed
Vakali, Athena
Terzi, Evimaria
Metadata
Show full item recordAbstract
Navigation has added interactivity in nowadays multimedia applications which support effective accessing to objects of various formats and presentation requirements. Storage issues need to be reconsidered for the new type of navigational multimedia applications, in order to improve system’s performance. This paper addresses the problem of multimedia data storage towards improving data accessibility and request servicing under navigational applications. A navigational graph-based model (for the multimedia data representation) is proposed to guide the data placement under a hierarchical storage topology. The multimedia data dependencies, access frequencies and timing constraints are used to characterize the graph nodes which correspond to multimedia objects allocated at the tertiary storage level. Based on certain defined popularity criteria, data are elevated and placed on secondary level towards improving both the request servicing and data accessibility. The proposed multimedia data elevation is a prefetching approach since it is performed “apriori” (not on demand) based on previously extracted user access patterns. Appropriate data placement policies are also employed at the secondary level, and a simulation model has been developed based on current commercial tertiary and secondary storage devices. This model is used to evaluate the proposed popularity based data elevation approach as employed under a hierarchical storage subsystem. Experimentation is performed under artificial data workloads and it is shown that the proposed hierarchical data placement approach considerably improves data accessing and request servicing in navigational multimedia applications. The iterative improvement placement is proven to outperform earlier related multimedia data placement policies with respect to commonly used performance metrics.
Collections
- Περιοδικά, εφημερίδες [1351]