An efficient cache management scheme for accessing small files in Distributed File Systems

Recently, the analysis of small files is required to provide individual users with the latest information and optimized services. In this paper, we propose a distributed cache management scheme that considers cache metadata for efficient accesses of small files in Hadoop Distributed File Systems (HDFS). The proposed scheme can reduce the number of metadata managed by a Name Node since many small files are merged and stored in a chunk. It also reduces unnecessary accesses by keeping the requested files using clients and the caches of data nodes and by synchronizing the metadata in client caches according to communication cycles.