计算机与现代化Issue(6)：19-24,29,7.DOI:10.3969/j.issn.1006-2475.2018.06.005
DoFFT:一种基于分布式数据库的快速傅里叶变换方法
DoFFT: A Fast Fourier Transform Method Based on Distributed Database
摘要
Abstract
Fast Fourier transform has a wide range of applications in astronomy. For example, pulsar signals often require coher-ent decler processing based on fast Fourier transforms. As the signal data is usually stored in the database, and executing fast Fourier transform algorithm outside of database after getting data from the database will have a lot of I／O and network overhead and thus seriously affect the overall performance. To solve this problem, this paper designs a DoFFT (Database optimized FFT) algorithm which can execute and optimize fast Fourier transforms in parallel in the form of user-defined functions (UDF). In ad-dition, different load of each node in the database cluster and uneven distribution of data may sometimes lead to inefficient imple-mentation, DoFFT method based on the CPU, I／O, network and transmission rates, and other costs, redistributes data to further optimize the parallel execution of fast Fourier transform. The experimental results show that the performance of DoFFT algorithm is improved effectively with the optimization based on data redistribution.关键词
分布式数据库/快速傅里叶变换/并行/数据分布/代价模型Key words
distributed database/fast Fourier transform/parallel/data distribution/cost model分类
信息技术与安全科学引用本文复制引用
季朋,李晖,陈梅,戴震宇..DoFFT:一种基于分布式数据库的快速傅里叶变换方法[J].计算机与现代化,2018,(6):19-24,29,7.基金项目
国家自然科学基金资助项目(61462012, 61562010, U1531246) （61462012, 61562010, U1531246）
贵州大学研究生创新基金资助项目(2017081) （2017081）
贵州省数据分析与云服务创新团队项目([2015]53) （[2015]53）
贵州省科技厅科技计划项目(LH[2016]7427) （LH[2016]7427）
