At the 2009 Very Large Databases Conference, Peter Boncz, Stefan Manegold, and Martin L. Kerstenand have been named as recipients of the VLDB "10 Year Best Paper Award" for the paper Database Architecture Optimized for the New Bottleneck: Memory Access.
The paper addresses the evolution of computer architecture and the problems that database programs face due to their legacy architecture hitting the so-called "memory wall." It describes new ways in which databases should be organized internally to optimize the use of CPU caches. The paper pioneered much of the subsequent research into data processing techniques that optimize the use of modern computer architecture, such as CPU cache and SIMD instructions. It has also brought new attention and recognition to column stores, specifically in the context of MonetDB, an open-source database system for high-performance applications in data mining, OLAP, GIS, XML Query, text and multimedia retrieval. MonetDB often achieves a speed improvement for SQL and XQuery over other open-source systems.
"I am very honored to receive this prestigious award, together with my CWI colleagues Martin Kersten and Stefan Manegold, which validates our work on the use of columnar data storage and in making database systems much more computationally efficient on modern hardware," said Boncz. "Our subsequent research, which led to VectorWise, has since further improved these techniques and coupled them with many new innovations that allow us now to scale such efficient query processing to terabyte scales on a single database server."
VectorWise is a spin-off of the database research team at the CWI Institute of Amsterdam. VectorWise is working exclusively with Ingres to bring this technology to market. The Ingres VectorWise project lets developers take advantage of advances in modern processor and storage hardware using SQL. It's believed that these developments will enable the next generation of business applications to integrate the instant analysis of vast amounts of business data with mission critical transaction processing. Data management technology based on the Ingres VectorWise project should allow businesses to manage their current data at highly reduced costs, or to scale up their database workloads to perform data analysis tasks that were previously not feasible. Initial results show that the Ingres VectorWise project achieves more than 10x performance gains.
