Stochastic soft real-time systems distinguish themselves from hard real-time systems in that there is no demand for every job in a task to meet its deadline. For these systems, provided that job deadline missing rate is below some threshold, total performance is acceptable. For a fixed- priority preemptive scheduled soft real-time system, we present probabilistic time demand analysis technique by which a lower bound on the probability that a job in a periodic task meets its deadline is estimated. We have also performed a simulation study of an example system and check the tightness of the lower bound. The simulation results shows PTDA has a high speed and an error below 10%. It can provide important reference for the design of soft real-time application.