| 基于拟态计算的大数据计算架构的研究与验证 |
投稿时间：2019-05-10 修订日期：2019-05-10  点此下载全文 |
| 引用本文： |
| 摘要点击次数: 197 |
| 全文下载次数: 0 |
|
|
| 中文摘要:为进一步提升通用大数据的计算效率，提出了基于拟态计算的大数据计算架构。架构将含有CPU、GPU、FPGA等多种计算结构的拟态计算体系融入大数据计算框架中，实现了上层大数据架构对于底层多种计算结构的按需动态调度，构建一个基于主动认知结构的、可根据计算场景灵活变换的、高效的拟态大数据计算架构，从而提供效率更高的计算方法。实验结果表明，基于拟态计算的大数据计算架构相比通用大数据计算架构，在数据维度达到亿级时，计算速率提升至少10倍，具有较高的计算效率。 |
| 中文关键词:拟态计算 加速计算 大数据 并行计算 Spark |
| |
| Research and verification of big data computing architecture based on mimic computing |
|
|
| Abstract:In order to further improve the computational efficiency of general big data, a big data computing architecture based on mimic computing is proposed. Architecture will contain the CPU and GPU, FPGA structure of simulated calculation system into big data computing framework, implements the upper large data structure for the underlying multiple computing on-demand dynamic scheduling of the structure, building a, according to the calculation based on the structure of active cognitive scene of flexible transformation, the highly effective mimic big data computing architectures, so as to provide more efficient calculation methods. The experimental results show that compared with the general big data computing architecture, the big data computing architecture based on mimic computing can improve the computing speed by at least 10 times when the data dimension reaches 100 million, and has higher computing efficiency. |
| keywords:mimic computing accelerated computing big data parallel computing Spark |
| 查看全文 查看/发表评论 下载pdf阅读器 |
次