NFSv4.1: New Features
Coping with dropped connections as well as client and server crashes is a major concern of distributed file systems. Though NFSv4 improved on earlier NFS versions, problems still remain. NFSv4.1 builds a session layer on top of the transport layer and not only solves many of the earlier problems, but does so in a simpler fashion.
NFSv4.1 extends the capabilities of its predecessors by supporting high-speed I/O to clustered servers, bringing about parallel I/O and increasing scalability and overall performance. Metadata, such as directory information, may be stored on different servers than data is, keeping it out of the data path. Metadata servers provide clients layout information allowing them to access data servers separately and in parallel.
Specialized protocols may be used to transfer data between clients and data servers. One such protocol, the NFSv4.1 file protocol, is described in the draft specification, but others such as OSD, an object protocol, and FC, a block protocol, are explicitly made possible. pNFS effectively bridges the gap between Network Attached Storage (NAS) and Storage Area Networks (SAN), giving NFS clients direct and parallel access to storage devices.
NFSv4 allows servers to delegate control of files to clients so that, in certain situations, clients may perform operations on files with minimal interaction with the server. NFSv4.1 extends this ability to other kinds of objects, particularly directories.