A distributed file system only really makes sense when several computers are connected to each other. The documentation published by GlusterFS states that at least three servers are required. However, the term “server” in this sense should not be taken literally. Virtually any kind of physical or emulated hardware can be integrated. Besides normal computers, the use of virtual machines is also feasible. This also comes with many benefits, especially with regard to flexibility.
Integrated servers act as nodes, which are connected to each other through the TCP/IP network. The integrated devices create a so-called trusted storage pool, whose memory is provided in the form of bricks. Volumes are then built from these bricks. These can subsequently be integrated and used like normal data carriers. Computers with access are identified as clients, but it is possible for one PC to be both a server and a client.
A special feature is the software’s tremendous scalability. Any number of nodes and bricks can then be added later on, and the size of the storage space can be adjusted according to any new requirements. The storage space to be managed has a maximum size of several petabytes.
In addition, GlusterFS guarantees reliability through redundancy. The risk of malfunction is initially distributed among several systems that can also be spatially separated from one another. It is also possible to set up RAID networks. However, in contrast to the standard specified distributed volume, a replicated volume must be stored in this case. As such, each file will be saved twice, which is called RAID mirroring.