The CDS was established in 1972 with the aim of collecting, selecting, evaluating and distributing information to the world wide astronomical community. The Catalogue of Stellar Identification (CSI) database was created from the early years. Later renamed SIMBAD, it quickly reached half a million objects. This represented a technical challenge on mainframes having 28 Mbytes hard disks. SIMBAD is more than just a collection of catalogues: from the beginning, it has associated cross-identifications and bibliography to objects. The database content is built by an experienced team of librarians and scientists. The evolution of the database, from CSI to SIMBAD 4, is presented, in particular from a hardware and software point of view. The main constraints have been long term maintainability, performances and real time updating performed by a distributed team. The current SIMBAD version (3), which has been operational for 15 years, will soon be replaced by a new version, based on current technologies, in particular JAVA and open source Database System, VO-enabled, ready again for future evolutions.