Database systems: Structure of a database file system for the UNIX operating system

A database file system is the low-level part of a database management system that handles data storage, indexing, concurrency control, transaction control, and recovery. This article describes a structure for a database file system that encapsulates important design decisions into separate modules. This permits a choice of implementations for each module. A database file system can be generated for a particular application by choosing the appropriate implementation for each module from a catalog (one from column A, one from column B, etc.). This approach can be both more flexible and more efficient than the usual approach of building a monolithic database file system to satisfy everyone. The challenge is to partition the database file system functions into modules that hide enough information to allow flexibility, but not so much as to be inefficient. The structure described here, which is biased towards information hiding, is an initial attempt at a suitable partitioning. An experimental system, XFS, was built to test this structure.