Object-based file system

In computing, the term object-based file system refers to an architectural blueprint for distributed file systems. An object-based file system splits file metadata (such as the filename, its size and access times) from file data and stores them on different servers. File metadata is stored on metadata servers, the file data is split into so-called objects and stored on object storage servers. The file system client employs metadata and object storage servers to present a full file system abstraction to the users.

The object-based file system architecture has several advantages:

• The file system client accesses file content directly via object storage servers. Metadata servers are only contacted once when the file is opened, thereby eliminating the metadata bottleneck of block-based distributed file systems.
• The size of these objects can usually be configured on per-file basis. This allows the stripe width to be adapted to the access properties of an individual file when the objects of a file are striped across multiple object storage servers.

File systems that follow an object-based architecture include Lustre and exofs.

Sources

Mesnier, M.; Ganger G., Riedel E. "Object-based storage". IEEE Communications Magazine 2003 8: 84–90. 

Factor, M.; Meth K. Naor D., Rodeh O., Satran J. (2005). "Object storage: The future building block for storage systems". Local to Global Data Interoperability - Challenges and Technologies. IEEE Computer Society. pp. 119–123. doi:10.1109/LGDI.2005.1612479. http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1612479.