Out From Under the Trees

This work is motivated by the observation that hard disk seek capability per byte of storage has declined exponentially for decades, and yet most popular access structures are still based on basic ideas proposed 30 years ago. These access structures are typically seek-intensive. In this paper, we propose a new paradigm for data access and organization called the linear file, which aims to drastically reduce disk seeks during query and update evaluation, by organizing data as a set of ongoing, on-disk sorts. In a linear file, updates are processed quickly, structures are 100% packed, and data are carefully organized on disk to reduce seeks among disk pages. Since the linear file is a general template for data organization, in this paper we benchmark T1SM and T2SM, two specific instantiations for use with one-dimensional and spatial data, respectively. 1. Contact author Paper ID: 417