Implementation And Optimization Of Client For Filesystem In Userspace

With the advent of the era of big data, digital information to be processed increases exponentially. Distributed file system provides an effective solution for big data storage and processing. Distributed File System can be divided into two categories according to the provided interfaces: one supports or partially supports the POSIX(Portable Operating System Interface) interface, while another provides proprietary interface. Most of the POSIX interface supported or partially supported Distributed File Systems chose to implement the client in kernel space for better performance.Developing in-kernel file systems for Unix is a challenging task, due to a variety of reasons. This approach requires the programmer to understand and deal with complicated kernel code and data structures, making new code prone to bugs caused by programming errors. Debugging kernel code is also tedious, and errors can require rebooting the system. Even a fully functional in-kernel file system has serveral disadvantages. Porting a file system written for a particular flaver of unix to a different one can require significant changes in the design and implementation of the file system. In contrast to kernel development, programming in user space mitigates, or completely eliminates, serveral of the abovementioned issues. FUSE is a widely used framework available for unix-like operating systems, that allows nonprivileged users to develop file systems in user space. In this paper, a userspace client for CAPFS(Cappella distributed file system) is designed and implemented. The program analyzes the access bottlenecks of userspace client and optimize the proformance by using userspace cache, RDMA(Remote Direct Memory Access), Direct IO etc.The test results show that the performance of userspace client come close to in-kernel client's, while dratically reducing development cycles.