Data Modeling for Better Performance in a Bulletin Board Application

Though alternatives of the use of relational databases are usually only recommended for very large collections of data, there are situations in which they may be convenient even for smaller datasets. An electronic notice board application, which allows members of the community it serves to join several groups, or only subgroups of the members, is used to illustrate such a situa-tion. The audience of a message also consists of certain groups, but may me limited to only some of their subgroups. Thus, at the start of a reading session by a member of the community, a query must determine the messages to be offered to the user. The situation called for a model, and the corresponding processing, that would efficiently build this list of messages to be presented to the user. Queries formulated in SQL, using a relational database model, failed to achieve satisfactory response times, even for relatively small collections of messages. So other models and processing modes were designed and tested, resulting in much better performance. A two stage query with the same model, where the second stage used the results of the first, but via programming code, though producing a thirtyfold reduction in response times, was still deemed insufficient. So a hy-brid model was formulated, where bitmaps were included as fields of an otherwise relational da-tabase, resulting in an additional twentyfold diminution of the duration of the query. For example, a query took a second using the hybrid model, instead of several minutes with the initial query. Simulated data for sets with different compositions of messages and their audiences were used to evaluate the performance of the models, and the resulting response times are compared and dis-cussed briefly, including comments on the use of relational and other database models. The mate-rials and concepts presented could serve as part of a course in data modeling