Heuristic algorithms for I/O scheduling for efficient retrieval of large objects from tertiary storage

Multimedia data service applications of today are to efficiently deal with a possibly massive amount of large objects (LOBs). The storage capacity of the traditional disk-based DBMS is certainly limited to support such applications. As such, it is necessary for the DBMS to employ tertiary storage devices, which perform often with long latency and yet can provide a huge amount of storage capacity at relatively low cost. We investigate tertiary I/O scheduling algorithms for efficient retrieval of LOBs that are stored in the tertiary storage device with multiple platters. Taking into consideration the performance characteristics of the tertiary storage devices and LOB retrieval, we propose a set of tertiary I/O scheduling algorithms, each of which employs a different heuristic for scheduling, reducing the latency involved in tertiary I/O for retrieving LOBs. The performance of the proposed algorithms is evaluated through detailed simulation experiments