 Abstract | The use of databases is a common practice for storing and organising large
amounts of data. In a distributed environment, especially widely distributed
systems, replication provides a means for storing an entire database locally. However,
in an environment where mobile devices are used, resource limitations on
both storage and power capacity eliminates the possibility of full replications.
Moreover, reliable access to centralised databases is not always possible in these
environments. Consequently, there is a need to find ways of storing data on a mobile
computer in such a way as to maximise user access while providing answers
at an acceptable level of accuracy.
This dissertation argues that it is useful to know and incorporate the user
context in which the database will be used in the creation process of the mobile
database. It investigates a context sensitive summarisation technique and
provides a proof of concept prototype, COSMOS, which maximises data availability
based on the context in which the system is being used and the accuracy
required by the user. More specifically, a framework for summarising data using
the context of the user is introduced.
Local nulls are proposed to the relational algebra and SQL to allow efficient
querying of summary databases. Some modifications to the database management
system include the handling of transactions in the system, update propagation
and system failure protocols. These modifications are proposed to allow the
effective use of the summarised data. |
