The technological gap between main and disk memory is constantly increasing, leading to a significant performance bottleneck for relational database management systems (RDBMS). This paper presents interesting research into the use of microelectromechanical systems (MEMS) as a base architecture for RDBMS. According to the authors, this approach is needed because of the technological gap.
A new flexible retrieval model (FRM) for MEMS-based storage devices is proposed. In particular, the authors exploit the physical characteristics of MEMS-based storage devices to develop a placement scheme for relational data that enables both row-wise and column-wise retrieval. As a result, they have developed algorithms for different relational operations based on such a data layout. Their experimental results and analysis demonstrate that their approach can result in significant improvements in relational databases. Additionally, the authors prove that their scheme improves not only input/output (I/O) utilization, but also cache performance for a variety of relational operations.
In general, this paper is well structured and organized. The authors provide a sufficient review of the existing data placement techniques for conventional hard disks, and describe their FRM for MEMS-based storage devices, as well as their processing techniques. Finally, they provide a performance evaluation of their FRM based on the Transaction Processing Performance Council benchmark H (TPC-H). It is possible that their work may have some impact on the early stages of architectural design of MEMS-based storage devices.
I strongly believe that this research would be very interesting and beneficial for all those involved in MEMS and RDBMS research.