A Comparative Benchmark of Large Objects

Originally Binary Large Objects (BLOBs) in databases were conceived as a means to cap-ture any large data (whatever large meant at the time of writing) which, for whatever reason, cannot or should not be modeled relationally. Today we find images, movies, XML, format-ted documents, and many more data types stored in database BLOBs. A particular chal-lenge obviously is moving such large units of data as fast as possible, hence performance benchmarks are of interest. However, while extensive evaluations have been undertaken for a variety of SQL work-loads, BLOBs have not been the target of thorough benchmarking up to now. TPC and SPC-2 standards do not address BLOB benchmarking either. We present a comparative BLOB benchmark of the leading commercial and open-source systems available under Unix/Linux. Commercial DBMSs are anonymised, open-source DBMSs benchmarked are PostgreSQL and MySQL. Measurements show large differences between the systems under test, depending on various parameters. A surprising result is that overall the open-source DBMSs in most situations outperform commercial systems if configured wisely. ACM Categories and Subject Descriptor