File Systems

Large-scale high performance computing systems generate, update, and retrieve enormous amounts of data over their lifetimes from various types of persistent storage. Without appropriate abstractions and associated user interfaces, governing and coordinating access to this information would not be practically feasible. File systems provide high-level semantics that hide the details of directly accessing storage devices at the block level. One of the most widespread input/output interfaces in use today is defined by the Portable Operating System Interface standard, discussed in this chapter along with usage examples. Parallel access to shared data in distributed environments introduces additional challenges that are addressed by distributed and parallel file systems. These are exemplified by the Network File System, the General Parallel File System, and the Lustre parallel file system.