Databases and data management form the core part of the current processing in the information systems. Data tuples are mostly delimited by the validity time frame. Nowadays, it is strictly necessary to store not only current valid data, but the whole evolution. Therefore, temporal paradigm has been created, which is based on the object identifier extension. Thus, the primary key of the object is extended by the time definition. This paper offers the cross-section of the temporality management, historical evolution and proposes new data models characterized by the various granularity of the processing. Our own proposed attribute oriented temporal model is based on storing any granularity and time spectrum, as well. Whereas some data columns are not necessary to be monitored over the time, even some of them do not evolve over the time at all, mentioned solution brings suitable solution. In the paper, we deal also with architecture improvements, which can manage synchronization groups. The first part of the paper deals with architecture of temporality. The second part evolves from the data management, background processes up to volatility and identification of the version. The third part proposes transaction definition and the impacts of the integrity).