The proliferation of parallel architectures demands the development of methods and the construction of tools that use these methods in the performance analysis prediction of applications (programs) running in architectures which are already in use or still being designed. A very important class of parallel architecture is called multiprocessors. This class allows the exploration of the parallelism that exist within the program`s code: programs are partitioned into smaller units (tasks) that execute concurrently in the processors of the architecture. This dissertation presents a tool - Tootisys - for modeling multiprocessors where processors share a global memory and where several programs execute at the same time. Programs are restricted to be CPU-bound and are represented by direct acyclic task graphs. The architecture is described by the time parameters of its processors and by a function that gives the delay in the access to the global memory. Tootisys allows the modeling of system where processors are disputed by tasks or programs and where multiple tasks can execute at the same time in each professor. The algorithm used by tootisys in the analysis of a concurrent system is iterative. This algorithm has pointed out to very computationally efficient: it has rapidly converged to final values in all cases studied. This dissertation also presents the graphic interface project of Tootisys and its implementation. Tootisys’s interface is a king of direct manipulation. However, some characteristcs of menu selection interfaces can also be found in it. The designed interface is efficient and minimizes error possibilities. Furthermore, it is easy to use, to learn and to remember. Since this interface was implemented in C and follows the X11 satandard, it is expected that Tootisys can be easily modifield to run in several machines.