CouchFS: A High-Performance File System for Large Data Sets

Numerous file systems have been implemented to meet the needs in today's big data era, however many of them require specific configurations or frameworks for data processing. This paper presents CouchFS, a POSIX-compliant distributed file system for large data sets. We build CouchFS on top of CouchDB, which grants us flexibility to handle semistructured data. Since a database has similar behaviors as a file system, and CouchDB provides a high customizable MapReduce view for indexing, CouchFS is able to achieve high-performance searching for both text and supported binary objects. This work compares search of Wikipedia data using CouchDB, PostgreSQL and Spotlight on HFS+ file system. We show our design of CouchFS and discuss future approaches to improve this file system.