A file system for accessing MySQL tables as CSV files

[摘要] In this thesis, we describe the design and implementation of a userspace file system that represents MySQL tables as comma-separated values (CSV) files. The users can access and modify the data through both the file system and MySQL;;s query interface. In order to transform read and write operations to SQL queries, we maintain a reverse index from file offsets to line numbers. Changes to the database outside of the file system are reflected on the file system by means of MySQL;;s master-slave replication feature. We evaluate our system by comparing its performance to a regular file system;;s performance using popular command line tools(grep, sed, awk) and user applications (OpenOffice.org spreadsheets). The choice of the database for this system was MySQL because of its popularity and the availability of its source code, however, the same ideas can be applied to any relational database with replication to create a similar system.

[发布机构] Massachusetts Institute of Technology