Parallelism in Database Operations
Helsinki University
Helsinki University, 2012
@article{karkkainen2012parallelism,
title={Parallelism in Database Operations},
author={K{"a}rkk{"a}inen, K.},
year={2012}
}
The developments in the memory and hard disk bandwidth latencies have made databases CPU bound. Recent studies have shown that this bottleneck can be helped with parallelism. We give a survey of the methodologies that can be used to implement this kind of parallelism. Mainly, there are two kinds of parallel disciplines that have been discussed: 1) CPU parallel; and 2) GPU parallel. The CPU parallel means employing the vector processor in the CPU. This is done by using certain machine code operations, that allows CPU to operate on multiple data items in the same instruction cycle. The GPU parallel means using the GPUs general programming interfaces to work on data in parallel. We review recent research that has been done in both these areas of database parallelism.
December 2, 2012 by hgpu
