Performing joins without decompression in a compressed database system

There has been much work on compressing database indexes, but less on compressing the data itself. We examine the performance gains to be made by compression outside the index. A novel compression algorithm is reported, which enables the processing of queries without decompressing data needed to perform join operations in a database built on a triple store. The results of modelling the performance of the database with and without compression are given and compared with other recent work in this area. It is found that for some applications, gains in performance of over 50% are achievable, and in OLTP-like situations, there are also gains to be made.