Multi-user Index-based Data Access System(MIDAS) has been developed as a storage subsystem for relational database systems. In this thesis, we enhance MIDAS in the following three aspects. First, we extend the volume management capability so that MIDAS can manage multiple volumes dynamically. Second, in order to support multimedia data, we design and implement a scheme that manages long data items and variable-length records. Finally, we improve the recovery method of MIDAS by implementing ARIES, which is known as an efficient recovery scheme. Through this work, MIDAS become a flexible storage subsystem that can be used in various applications.