Energy Efficient Mobile DB

Abstract

In this work, we propose Database Shadowing, or DASH in short, which is a new crash recovery technique for SQLite DBMS. In the current journal mode, there are two problems: write amplification problems and excessive fdatasync() for order assurance. Both problems cause performance degradation and serious energy consumption. Various studies conducted to solve the two problems. However, the results of the studies conducted are a lot of code, low in compatibility and require additional hardware. We created a (LoC 90) Metadata Embedding that modifies the current PERSIST journal mode in commercialization. Based on Insert mode, Metadata Embedding improves performance by approximately 168% compared to conventional PERSIST journal mode. We also reduced the power consumption of one transaction by 9 mJ / Tx.