On Configuring Hierarchical Storage Structures

The rapid progress in mass storage technology is enabling designers to implement large databases for a variety of applications. The possible configurations of the hierarchical storage structures (HSS) are numerous and result in various tradeoffs, e.g., storage cost, throughput, initial latency, etc. A naive design might waste system resources and result in high cost. For example, one naive design decision is to add disk caches for those applications whose bandwidth is already satisfied by the tertiary storage device and can tolerate a high latency. In this study, we introduce a configuration planner for HSS in support of video-ondemand applications. The input to the planner are the number of simultaneous displays required by the target application (i.e., throughput), database size, and expected pattern of access to the objects. It’s output are the choice of mass storage devices at each level of the hierarchy. The resulting configuration supports the application requirements at a minimum cost per display. It is possible to extend this study to consider other application requirements, such as: initial latency and storage reliability. Moreover, the presented concepts can be generalized to scientific applications, such as those