An NVM-Aware Storage Layout for Analytical Workloads

Abstract

As DRAM is reaching its scalability limit, Non-Volatile Memory (NVM) technologies are moving more and more into focus to meet the requirements of modern database and Big Data systems. With this novel paradigm a lot of new opportunities but also challenges emerge which requires a rethinking of database system developers regarding the programming model and data placement. In this paper, we present a design of an NVM-aware storage layout for tables using a multi-dimensional clustering approach and a block-like structure to utilize the complete memory stack. Contrary to previous works, this approach focuses on analytical workloads. In a microbenchmark, its potential for scan and point queries is investigated. We show that our approach is able to outperform the considered competitors by up to one order of magnitude.