Modularized Redundant Parallel Virtual File System

Providing data availability in a high performance computing environment is very important, especially in this data-intensive world. Most clusters either use RAID technology or redundant nodes to ensure data availability. People usually use parallel file systems to increase the throughput of a computing system. However, when a parallel file system is involved in a distributed environment, some mechanisms must be provided to overcome the side-effect of using striping. PVFS is a popular and open source parallel file system in the Linux environment, but it provides no fault tolerance. We propose an idea of using distributed RAID technology to ensure the data availability of using striping. By introducing a parity cache table (PCT), we can improve write performance when updating parity is needed. The evaluation of our MRPVFS (Modularized Redundant Parallel Virtual File System) shows that the read performance of MRPVFS is almost the same as that of the original PVFS. As to the write performance, there only exists a little performance degradation due to generating parities.