Architecture of a Distributed File System

The internal design of a distributed file system is built around one big idea: keep the information about files separate from the files themselves. The information about a file, such as its name, its size, its permissions, and where its blocks live, is called metadata. The actual contents of the file are just data blocks.

Because of this split, most distributed file systems have three main roles. Clients ask for files. A metadata server answers questions about files. Storage nodes hold the actual blocks. Details vary between systems, but this shape appears again and again.