Memory-Based Data Storing Technologies on Hadoop Distribution File System
Aibo Song, Jinlin Tu, Jinghua Zhao, Xuejiao Qian · 2015
A single storage medium of mass data storage lead to low throughput and long data access time delay. To solve these problems, Memory-based Data Storing Technologies on Hadoop Distribution File System is related. First, the calculation of memory available on each storage node, this part of the memory mapped to disk, and realize the HDFS to this part of the effective management of memory resources. And to establish a reasonable memory data exchange model, when the data is stored in the memory load growth to a certain extent, the memory file heat low displacement to disk, disk file in high heat move back into memory. Comparing with the HDFS based on disk storage, in this paper, the HDFS based on the memory data storage system can effectively shorten data access delay, improve the system of data throughput, with significant performance advantages.