New Hardware Support for Database Operations Nagender Bandi, Chengyu Sun, Hailing Yu, Divyakant Agrawal, and Amr El Abbadi   View Paper (PDF)   Return to June 2005, Databases for New Hardware, Volume 28, Number 2 Abstract Query processing cost in database applications can be divided into two parts: the I/O cost and the computational cost. Traditionally DBMS researchers have focused on the issue of reducing the I/O cost as it is the primary source of bottleneck in most of the database operations. As the computer systems which run the DBMSs evolved, this problem worsened because the speeds of secondary storage devices increased at a much slower pace compared to the memory access and processing speeds. Further, as databases become increasingly accepted in areas such as spatial databases, we encounter new challenges. For example, the computational cost has been known to be the bottleneck in certain spatial operations. In this paper, we present a survey of our work where we present novel solutions which use hardware to alleviate the I/O cost bottleneck for traditional database applications and the computational cost bottleneck for non-standard database applications such as spatial databases. In particular, we use the novel MEMS-based storage devices to develop efficient data placement solutions for relational databases and two-dimensional datasets. Using commodity graphics hardware, we develop novel spatial solutions which outperform conventional software techniques. We also integrate these techniques into a popular commercial spatial database and demonstrate significant improvement in query performance. ©2006 Association for Computing Machinery