There are two major requirements in real-time data processing systems. The first one is to guarantee that the system can handle its peak processing load without losing any input real-time data from the devices. The second one is to guarantee that the worst case response times of all the real-time jobs in the system should be less than the required upper bound response times of them. In the literature, Joseph and Pandya's analytic model was suggested for performance evaluation of real-time system. But the model has an assumption that jobs use only CPU, so it cannot be applied to the real-time data aquisition storage and retrieval systems in which jobs use not only CPU but also disk.
In this thesis, a simulation model is proposed for the real-time multiprogramming systems in which each job uses not only CPU but also disk. To prove the correctness of the model, benchmark programs which represent the jobs were run on VAX 8200, and the response times of them were measured and compared with the predicted results by the simulation model. The simulation results showed that simulation error were within 25 percent when the worst case response time was overestimated, and within 15 percent when it was underestimated.