Stored procedures are an important feature of all major database systems that allows to execute application logic within database servers. This paper reports on experiences to implement a popular scientific algorithm, the Basic Local Alignment Search Tool (BLAST), as stored procedures within a relational database. We implemented the un-gapped, nucleotide version of the BLAST algorithm with four different relational database engines, both commercial and open source. In an experimental

more »

... we compared our dbBLAST implementations with a standard file-based BLAST implementation from NCBI with regard to the implementation effort, runtime performance, and scalability. It shows that although our dbBLAST runs faster than the file-based BLAST program for short query sequences, all implementations lack scalability. However, the results also indicate that stored procedures require significant less development effort-both in time and space-than traditional programming approaches.