Abstract: The present invention relates to a method to of ensuring the reliability of a system distributed data base, comprising several computers which together form nodes (1, 2, 3, 4) wherein information belonging to the distributed data base is storable. A fragment comprises a part of the data base and includes a primary replica (11) of the part and a secondary replica (21) of the part. The primary replica (11) is stored within a first node (1) and the secondary replica (21) is stored within a second node (2), separated from the first node (1). Both data information and log information are stored within both the primary replica (11) and the secondary replica (21). The secondary replica (21) is updated when the primary replica (11) changes (A) and is used to re-create a primary replica should the first node (1) crash. Respective fragment also comprise a stand-by replica (31) of the part, which is stored in a third node (3) that is separate from the first (1) and the second (2) node.