Database system

A database system, comprising a plurality of computing devices (4) and an allocator (6) which is designed to distribute the data of the database over the computing devices (4), wherein the computing devices (4) are each designated by a host index and each comprise a random-access memory,

• wherein the allocator (6) is configured:
• to organize a plurality of spaces indexed over each random-access memory, and to organize a part at least of the data of the database into tranches having a tranche index,
• to define a plurality of mutually distinct sequences each comprising a plurality of pairs whose number is equal to the number of computing devices (4), each pair associating a unique rank value and a host index,
• to maintain a list of the functional computing devices (4) over time,
• to associate each tranche with a sequence of the plurality of sequences, on the basis of the list of functional computing devices (4), so that, for two tranches having respectively a first tranche index and a second tranche index such that the difference between the first tranche index and the second tranche index is equal to the number of sequences of the plurality of sequences, a given sequence of the plurality of sequences is associated with a single tranche whose tranche index lies between the first tranche index and the second tranche index,
• to define, for a tranche of given tranche index, from among the list of functional computing devices (4), a master space on the computing device (4) whose host index is associated with the lowest rank in the sequence associated with the given tranche, the master space being chosen on the computing device (4) as the space whose index corresponds to the given tranche index,
• to define, for a tranche of given tranche index, from among the list of functional computing devices (4), a plurality of slave spaces, chosen on computing devices (4) in such a way that the sum of ranks which are associated with the host indices of these computing devices (4) in the sequence associated with the given tranche is a minimum, each slave space being chosen on the respective computing device (4) as the space whose index corresponds to the given tranche index, and the computing devices (4) are configured such that one of the computing devices (4) which receives a master space replicates the data of the associated tranche on the corresponding slave spaces.