It's a guarantee that when a file is closed subsequent opens will see the latest changes. This is achieved by the client flushing all cached changes to the server when it closes the file, and clients opening a file must ignore all cached info they have about the file. This is as opposed to time-bounded consistency where, even after a file is closed, it may take a period of time before the server and subsequent clients see the changes.
Traditional NFS has write-through caching. When a file is closed ALL modified blocks are sent to a server. The close() function does not return until each byte in each block is safely stored. However if two clients on a network are writing to a cached copy of the same file and issue a close command in close proximity, NFS can't gaurantee whose modified data will make it to master copy. To avoid this scenario close-to-open cache conistency prevents two or more clients from modifying a file simulateneously. Every client must query a server before modifying it. This trades off scalability with consistency.
非共享;
Close before Open;
WriteBack,缓存;