Using research funded in part by Intel, Ingres has released a newly re-engineered version of its open-source database management system, called Ingres VectorWise.
The company claims that this database engine, when used for analysing large data sets, can run 10 to 70 times faster than standard SQL relational database systems, based on beta versions run by customers.
Amsterdam research institute CWI originally developed the underlying database engine, called VectorWise, spinning off a company of the same name to commercialise the technology. Intel helped fund development of the technology.
The performance gains come about thanks to changes in how the query processing works, said Roger Burkhardt, CEO of Ingres. The architecture takes advantage of features in Intel Nehalem-based processors, such as out-of-order execution capabilities, chip multithreading and larger L2/L3 caches.
The software also uses novel compression algorithms and data-storing procedures. "VectorWise data decompression can make queries that would otherwise become I/O bound significantly faster. It can even accelerate parallel queries where multi-core CPUs would otherwise become RAM bandwidth starved," states a white paper on VectorWise.
The Vectorwise database works with standard SQL, according to Burkhardt. The software, which runs on Linux, is available for download.