Distributed file systems and distributed memory

The model of a single file system shared by all users of a computer is not only convenient but expected by most computer users. It seems natural to extend this model across multiple computers so that all users on a collection of computers share the same file system, thus forming a distributed file system. Similarly, the model of a collection of threads of control sharing the same address space as they cooperate in a computation is attractive for exploiting concurrency. This single-address-space abstraction is certainly the natural model for use on a shared-memory multiprocessor. Its convenience for programming is so compelling that it is used increasingly to take advantage of parallelism on distributed systems, where it is called distributed memory.