Semantic Database Compression System Based on Augmented Vector Quantization

In the last years, that amount of data stored in databases has increased extremely with the widespread use of databases and the rapid adoption of information systems and data warehouse technologies. It is a challenge to store and recover this increased data in an efficient method. This challenge will potentially appeal in database systems for two causes: storage cost reduction and performance improvement. Lossy compression in databases can return better compression ratios than lossless compression in general, but is rarely used due to the concern of losing data. For relational databases, using standard compression techniques like Gzip or Zip don't take advantage of the relational properties; since these techniques don't look at the nature of the data. In this paper, we propose a database compression system that takes advantage of attributes semantics and data-mining models to find frequent attribute pattern with maximum gain to perform compression of massive table's data. Furthermore, the suggested system relies on augmented vector quantization (AVQ) algorithm to achieve lossless compression version without losing any information. Extensive experiments were conducted and the results indicate the superiority of the system with respect to previously known techniques.