Journals Proceedings

Abstract

Solid State Disks (SSDs) are promising data storage devices in term of performance and energy consumption comparing to Hard Drive Disks (HDDs). They are more and more used, even in dataintensive applications where hard constraints are put on throughtput and response time. Accessing data in parallel becomes capital to satisfy these performance requirements. The SSD internal structure provides a potential for parallel access at different levels, up to its dies and planes. However, this parallel access relies completely on the data layout. In this paper, a data placement algorithm is presented. It distributes data across the SSD components and maintains this layout after write operations. This makes the parallel requests execution possible at the lowest levels. The impact of this data layout on the reduction of the waiting time is evaluated and the obtained results show a significant gain for OLTP and scientific applications up to a factor of 10.