This paper describes the Oz Scheduler, a workbench for scheduling problems. Through a graphical interface, the user can freely combine the
elements that define a scheduling strategy. Such elements include
constraints with different popagation behavior or distribution and
search strategies. Exploring the possible combinations can lead to
better solutions. Recent and successful techniques for scheduling are
incorporated.
Resulting from the selections made, a constraint problem is generated
dynamically. For this problem, the solution, statistics and the search
can be inspected by several graphical tools. The functionality can be
extended by sending messages to the Scheduler. The functionality and
the implementation of the Oz Scheduler are discussed. The overall
performance of the Scheduler for standard benchmarks is comparable to
state-of-the-art special-purpose systems for scheduling. The
implementation is based on the concurrent constraint language Oz.