Abstract: A method of assigning transaction identifiers assigns an identifier that is unique to each transaction, each transaction operating on data distributed across various stations. The method determines a first field of the identifier, having either a temporary value identifiable as such and unique to the station executing the application process that requested opening of the transaction if opening of the transaction is requested during an initialization phase or a value assigned by a first manager process specific to the station executing the requesting application process if opening of the transaction is requested after the initialization phase. The method also determines a second field having either a temporary value identifiable as such and unique to the set of stations if the opening of the transaction is requested during the initialization phase or a value assigned by a second manager process common to the set of stations if opening of the transaction is requested after the initialization phase.

Abstract: A transactional protocol allocates write intention locks at the time of access to a transaction modifying data stored in a memory. These locks do not prohibit the allocation of read locks to other transactions. The other transactions can therefore read the old values of the variables until the new values are written into the memory when the transaction is closed. The time for which reading of data being modified is prohibited is restricted to the time to write the new values into the memory. This protocol is of particular benefit if the memory is a volatile memory and the modified data is backed up in another, non-volatile memory. In this case, the protocol further keeps up-to-date a table stored in the non-volatile memory and containing the identifiers of current transactions. This table is used to ensure that the data is consistent after failure of the volatile memory and contributes to maintaining a minimal time of prohibition of reading of data being modified.