Title :
Research of Distributed Data Store Based on HDFS
Author :
Xin Wang ; Jianhua Su
Author_Institution :
Dept. of Comput. Sci., China Univ. of Pet., Beijing, China
Abstract :
This paper analyzes the advantages and disadvantages of the traditional HDFS cluster architecture. An application framework designed for the small and medium-sized HDFS clusters is proposed in this paper to make up for incomplete HDFS application architecture, difficulties on deploying, inefficiency on I/O processing of small files. In addition, a caching mechanism is added to optimize the operation system on processing small files. Experiments show that this plan is feasible and the increased disk cache can significantly improve the efficiency of processing small files and optimize the system.
Keywords :
data handling; distributed processing; file organisation; HDFS application architecture; HDFS cluster architecture; Hadoop distributed file system; I/O processing; caching mechanism; distributed data store research; operation system; Buffer storage; Computer architecture; Distributed databases; File systems; Instruction sets; Internet; Servers; Cache; HDFS; Small Files I/O;
Conference_Titel :
Computational and Information Sciences (ICCIS), 2013 Fifth International Conference on
Conference_Location :
Shiyang
DOI :
10.1109/ICCIS.2013.384