DHFS: A High-Throughput Heterogeneous File System Based on Mainframe for Cloud Storage

Abstract

The file system for the Cloud Storage usually use the distributed structure which store the metadata and the user data respectively. The I/O nodes get the management information of file from the specific metadata server through the networks. Under this mode, the metadata server should be the performance bottleneck of the whole system when huge number nodes or the heavy workloads in Cloud Storage. Therefore, the capability of the metadata server become one of the key aspect for the whole system. In this paper, we present DHFS(Distribute Heterogeneous File System), a high throughput system for Cloud Storage, building metadata server on the high performance mainframe. For the different organization of the file system between mainframe and open system(Unix, Solaris, Windows, Linux, etc.), we implement the user-level library for the heterogeneous share of data which allow application in open system accessing files of mainframe. We also employ a series of strategies for the performance optimization.