The construction of large scientific experiments (such as the Compact Muon Solenoid experiment being undertaken for CERN) necessitates the use of complex production management operations, often distributed over many geographically separated research institutes. These workflow operations are often only loosely-defined at the outset of the construction and can evolve rapidly as the results of experiments are analysed. Existing commercial workflow management systems are largely based on relational database management systems (RDBMSs) and are unable to cope with the requirements of such scientific workflow environments and, in particular, with the dynamic schema evolution which results from the rapidly evolving scientific workflow definitions. This paper reports on the requirements for object repositories in the implementation of a prototype scientific workflow management system entitled CRISTAL (Cooperating Repositories and an Information System for Tracking Assembly Lifecycles) and considers issues surrounding data duplication between object repositories and scientific workflow management in CRISTAL.
展开▼