Mentor
Alférez, Germán
Document Type
Thesis
Publication Date
Spring 4-19-2024
Abstract
A routine scenario when developing PC applications is storing data in small files or records and then retrieving and manipulating that data with a distinctive identifier (ID). In these scenarios, the developer can save the records using the ID as the filename or use an embedded on-disk key-value database. However, many file systems can have performance issues when handling many small files. As a result, developers would rather avoid depending on an embedded database if it offers little benefit or has a detrimental effect on performance for their use case. Our contribution is to compare several key-value databases—SQlite3, LevelDB, RocksDB, and Berkeley DB—based on many parameters, including the file system—NTFS, as opposed to ext4, the file system utilized on a previous project \cite{pastpaper}—and explain the outcomes. Moreover, the metrics and technologies to be evaluated extend the metrics evaluated in our previous research work. We compare these key-value databases on two machines: a solid-state drive and a hard disk drive. Our research used the Windows Subsystem for Linux 2 (WSL 2) to work with.
Recommended Citation
Cunningham, Nicholas, "A Further Performance Comparison of Operations in the File System and in Embedded Key-Value Databases" (2024). MS in Computer Science Project Reports. 12.
https://knowledge.e.southern.edu/mscs_reports/12