Designing Distributed Databases from an Organisational Perspective Ant´ onio Rito Silva, Helena Galhardas, Paulo Sousa, Jorge Silva and Pedro Sousa IST/INESC, R. Alves Redol n 9, 1000 Lisboa, PORTUGAL Tel: +351-1-3100287 – Fax: +351-1-525843 email: Rito.Silva@inesc.pt, Helena.Galhardas@inesc.pt Abstract In this paper we present a case study of the database system design in the context of an organisation with special emphasis on the replication concern. This case study allows us to identify abstractions and concepts of organisational modelling which are reusable in similar cases. We used the TASKON/OOram method and tool to support the case study development. Keywords: Methodologies and modelling for CoopIS, Systems Integration and Interoperability. 1 Introduction Perceiving organisations in terms of their information systems is a common attitude. The information sys- tems structure, often called schema, gives a partial description of the organisation, which integrated with other such descriptions, is held to be the organisation model. This bottom-up approach for perceiving or- ganisations has been proven to be unsuitable to meet the new challenges faced by organisations: improve- ment and change in order to constantly adapt to new environment requirements. The bottom-up approach does not offer a global view of organisations where changes can easily be applied. Thus, a top-down view of an organisation is required. In this view, informa- tion systems are the result and not the cause of the organisation model. The organisation model, defined independently of the information systems, permits a global descrip- tion of the organisation, where management requests apply using the correct level of abstraction. Moreover, the organisation model is the reference when integrat- ing other information systems in the organisation. The design of information systems in general and database systems in particular can take advantage of an organisational perspective. Once the global organisation model has been defined, new database schemas should be designed and integrated as parts of the global one. Furthermore, the organisation mod- el permits the description of interoperability require- ments among databases since inter-database semantics is maintained. Organisations replicate information for two main reasons: circumstance and design. Replication by cir- cumstance happens, quite often, when the organisa- tion grows and changes in an uncontrolled way. It is locally decided when and how to replicate, based on subjective factors, e.g. communication between departments. The amount of replicated information brings the possibility of inconsistency. Nevertheless, replication is necessary for the normal behavior of an organisation. Replication by design corresponds to a conscious attitude where replication is driven by both computational and organisational factors. Tradition- ally, computational factors for replication are identi- fied as availability and reliability: availability permits access to replicated information without remote ac- cesses; reliability hides failures of remote databases from the local computation. Organisational factors for replication can be driven by geographic distribution, strategic decisions, security concerns, process tracing, and so on. Inside organisations, sometimes it is permissible that replicated information is not consistent. Incon- sistency can be allowed since most information is not critical and it is clear from an organisational perspec- tive where to find up to date information and, if it is the case, when to reestablish consistency. We call repli- cation models to the different perspectives an organi- sation has about its own replicated information. The replication models should be described within the or- ganisation model. This way, consistency of replication 1