A distributed file system is convenient only if several computers are connected to each other. According to the official description of GlusterFS, at least three servers are required , but these are not servers in the literal sense, but can be practically any type of physical or emulated hardware . Virtual machines can also be used, in addition to all kinds of computers, which have many advantages, especially in terms of flexibility.
Integrated servers operate as nodes (nodes) and connected through the TCP / IP network. The connected devices thus form what is called a trusted pool , that is, a set of trusted servers. These servers make their memories available to the user in the form of bricks (bricks) , in which volumes are finally created (units of volume) , which can then be incorporated and used as normal storage media. The computers that access the system are called clients or clients and it is possible that they also function as a server at the same time.
A special feature of GlusterFS is its great scalability , which allows as many nodes and bricks as desired to be added later . In this way, the storage space can be continuously adapted to the requirements of each moment. The maximum size of storage space that can be handled is several petabytes .
Additionally, GlusterFS ensures that data is not lost in the event of a crash by saving it redundantly . In this way, the risk is spread over several systems, which may be on separate physical media. It is also possible to create RAID-type clusters , which require that a copied drive ( replicated volume ) be added instead of a distributed one ( distributed volume ), as is usually the default option. This copied drive saves each file in duplicate and corresponds to the so-called mirror RAID .