Managing databases with binary large objects

Abstract

We present recommendations on Performance Management for databases supporting Binary Large Objects (BLOB) that, under a wide range of conditions, save both storage space and database transactions processing time. The research shows that for database applications where ad hoc retrieval queries prevail, storing the actual values of BLOBs in the database may be the best choice to achieve better performance, whereas storing BLOBs externally is the best approach where multiple Delete/Insert/Update operations on BLOBs dominate. Performance measurements are used to discover System Performance Bottlenecks and their resolution. We propose a strategy of archiving large data collections in order to reduce data management overhead in the Relational Database and maintain acceptable response time.