A key problem in providing 'enterprise-wide' information is the integr
ation of databases that have been independently developed. An importan
t requirement is to accommodate heterogeneity and maintain the autonom
y of component databases. Myriad is a federated database prototype dev
eloped at the University of Minnesota, to provide a testbed for invest
igating alternatives in architecture and algorithms for database integ
ration, query processing and optimization, and concurrency control and
recovery. The system incorporates our group's research results in the
se areas. This paper describes our experiences in the design and imple
mentation of Myriad, and in the project management. Special emphasis i
s given to discussing design alternatives and their impact on Myriad.
This paper also presents the software engineering principles and the p
roject management techniques we used in developing Myriad and the less
ons we learned. We believe these lessons would be useful for practitio
ners who wish to develop a similar system. Handling heterogeneity and
autonomy were prime objectives throughout the prototyping effort. We a
re convinced that a prototype federated database is an important infra
structural requirement for the overall goal of 'enterprise-integration
', and believe Myriad to be a significant contribution towards this.