Data Processing in the Modern Hardware Landscape

Download or read book Data Processing in the Modern Hardware Landscape written by Yannis Chronis and published by . This book was released on 2022 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Enterprises are increasingly relying on data-driven decisions in nearly every aspect of their business functions, and databases are at the center of the computing infrastructure that enables this capability. Naturally, the demand for faster analytics continues to grow unabated. Database engineers and researchers rely on effectively harnessing the power of hardware to achieve high data processing performance. In recent years, the hardware landscape has been changing drastically as a result of two key technological trends. Adapting data processing algorithms and systems to the new hardware reality is crucial in order to meet the processing needs of the future. The first key technological trend is termed the ``Memory Wall'', a name that captures that processors are being increasingly bottlenecked by memory performance in traditional hardware architectures. We propose new algorithms for the core data operation of searching sorted data and the class of top-K queries with large outputs. The proposed algorithms are designed around modern processing needs and the modern hardware performance characteristics as they are shaped by the "Memory Wall''. The second key technological trend we study is the result of Dennard's scaling breaking down. We will soon reach the limit of being able to improve the performance of processors by simply scaling down the manufacturing scale. Novel architectures and emerging hardware technologies are being studied as solutions to keep improving performance. In the last chapter of this thesis, we take a codesign approach to develop a database software system and an associative processor for analytic workloads. Associative processors have re-emerged as an attractive architecture that offers very large data-level parallelism. We show that following a codesign approach associative processors can be used to accelerate analytic query processing.