pNFS removes the performance bottleneck in traditional network-attached storage (NAS) systems by allowing the compute clients to read and write data directly and in parallel, to and from the physical storage devices. The pNFS architecture also eliminates many of the scalability and performance issues associated with NFS servers because file metadata and protocol data are transmitted over the Internet Protocol (IP) network, while file data travels over the storage architecture as files, blocks or objects.
One benefit of a parallel NFS-based system is that application servers, or clients, can gain simultaneous access in parallel over multiple data paths to storage servers or nodes. A metadata server out of the data path supplies the client with the location of the data. The client can then read and write data directly to the storage.
According to the Storage Networking Industry Association (SNIA), workloads with many small files or very large files that run on compute clusters could benefit the most from the simultaneous, parallel data access provided by pNFS-based systems. The first pNFS prototype was written by Dean Hildebrand at the University of Michigan in 2004.