Performance Comparison on the Heterogeneous File System in Cloud Storage Systems

The number of small- and mid-cap companies increases in recent decades due to the change on technologies. A considerable number of information systems are thus built for efficient production and management. As a result, a large number of operations and data are generated. The internal IT unit must have the ability to process the large amount of data, provide different services for each kinds of users, and make proper data backup. Because of the limited human resources and budget, a self-owned cloud storage system, that can effectively unify the existing and future storage system, may be a good choice to fulfill the needs. Functions that should be considered in the cloud storage system include: storage resource virtualization, construction of storage pools, supply of online data access, and backup of web services. In this study, we compare three common file systems, namely GlusterFS, Lustre, and Hadoop, on the time cost and efficiency as well as build up a cloud storage system combining these three file systems. Experimental results are as follows. First, the Hadoop has the best performance when the storage is of larger than 200MB. Second, the Lustre performs the best when the storage is of less than 200MB and more than 4MB file. Last, the GlusterFS is the best when the storage is of less than 4MB.