Hardware Partitioning for Big Data Analytics

Author

Wu, Liang ; Barker, Raymond J. ; Kim, Martha A. ; Ross, Kenneth A.

Author_Institution

Comput. Sci. Dept., Columbia Univ., Santa Clara, CA, USA

Volume

34

Issue

3

fYear

2014

fDate

May-June 2014

Firstpage

109

Lastpage

119

Abstract

Targeted deployment of hardware accelerators can improve the throughput and energy efficiency of large-scale data processing. Data partitioning is a critical operation for manipulating large datasets and is often the limiting factor in database performance. A hardware-software streaming framework offers a seamless execution environment for streaming accelerators such as the Hardware-Accelerated Range Partitioner (HARP). Together, the streaming framework and HARP provide an order of magnitude improvement in partitioning and energy performance.

Keywords

Big Data; data analysis; energy conservation; power aware computing; Big data analytics; data partitioning; database performance; energy efficiency improvement; hardware accelerators deployment; hardware partitioning; hardware-software streaming; large-scale data processing; range partitioning; throughput improvement; Accelerators; Bandwidth; Databases; Energy efficiency; Hardware; Large-scale systems; Servers; Throughput; Accelerators; Bandwidth; Databases; Energy efficiency; Hardware; Large-scale systems; Servers; Throughput; accelerator; big data; data partitioning; hardware; microarchitecture; specialized functional unit; streaming data;

fLanguage

English

Journal_Title

Micro, IEEE

Publisher

ieee

ISSN

0272-1732

Type

jour

DOI

10.1109/MM.2014.11

Filename

6762799