|
The topic of this thesis is the efficient implementation of the
bitemporal timeslice query. It answers queries of the type: What
knowledge about the state of the world at valid time, vtime, was
current in the database at transaction time, ttime? Differential
timeslice computation takes an existing cached timeslice as the
outset and applies necessary changes to get the result of the new
timeslice query. This is efficient because two timeslices with
similar valid time and transaction time can be expected to have a
large number of tuples in common. Differential computation has
previously been applied to transaction-time databases. This scheme
is extended to suit bitemporal data bases. In addition, the
algorithms are generalized for history timeslices that cover a
transaction or valid time period, and to n-dimensional timeslices
for applications with higher dimensions of time. Some aspects of
physical data storage in an implementation of the algorithm are
discussed, the complexity is analyzed and compared with existing
solutions. Directions for future research are given.
|