Journal of Systems Integration, 4, 151-169 (1994) 9 1994 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands. Quantification of Autonomy on Multidatabase Systems KEN BARKER BARKER@CS.UMANITOBA .CA Department of Computer Science, University of Manitoba, Winnipeg, Manitoba, Canada R3T 2N2 (Received March 16, 1992; Revised July 7, 1993) Abstract. A multidatabase system is an interconnected collection of autonomous databases each managed by an autonomous database management system (DBMS). When integrating multiple DBMSs, the key is the autonomy of the underlying participants. Much research has been undertaken in the past five years aimed at describing and building an integrated multidatabase system, but to date the term autonomy has only been defined intuitively. This article provides a rigorous definition for autonomy tailored to the multidatabase environment specifically but applicable to any system environment that involves the collaboration of autonomous participants. The major contribution of this article is a technique that measures autonomy along multiple dimensions so a single numeric value describing the amount of autonomy violated by a particular system design is quantified. This has a two-fold implication. First, the technique described forces researchers to consider autonomy from several different aspects that may not be the central focus of their research, but must be considered because assumptions made regarding one aspect of a system may have implications in other areas. Second, the value can be used as a measure for direct comparison among different systems or proposals. Finally, the article demonstrates the quantification tech- nique's applicability by applying it to several recent multidatabase research efforts. Key Words: Multidatabases, autonomy, multidatabase management system. 1. Introduction A multidatabase system (MDS) is an interconnected collection of autonomous databases. These autonomous systems must be integrated to form a multidatabase management system (MDMS). Such systems have been receiving increasing interest as the foundation of in- teroperable application environments. The database community's interest in interoperability has historically been in the context of heterogeneous databases. The most important characteristic of an MDS is autonomy. A number of important operational implications result, however, current research efforts, both theoretical and applied, have made different assumptions about precisely what autonomy means. Recently it has become increasingly apparent that a more rigorous definition of autonomy, suitable for multidatabase environ- ment, is needed before an in-depth discussion comparing current systems can be under- taken. Without rigorous definition it is likely that assumptions made when undertaking system integration will lead to systems that assimilate their participants rather than integrate them. This is a problem because when a system is assimilated it becomes part of the large multidatabase system per se thereby giving up all of its autonomy. An integrated system is one where the participants exhibit a great deal of autonomy but permit users to exchange their data.