Optimal Layout Scheme for Compound Objects in Multiple Object-based Storage Tape Drives

In a large scale of mass data management, tape library system has recently aroused lots of attention in terms of improvement of response time with the broad application of its backup usage. In this work, we address the optimal placement scheme of data object in parallel tape drives. By observing a pattern of request containing the repeated group of data objects, we propose a simple relationship measure between two data which a clustering algorithm based on replication can be conducted with. In order to reduce the contention among drives highly dependent on traffic loads which may occur in allocating a group of data into multiple tapes, we suggest a layout scheme which objects in a cluster is constrained by the number of tapes to be allocated while increasing the utilization of tape drive power.