Data grid systems are utilized to share, manage, and process large data sets. On the other hand, an increasing number of applications with real-time constraints arise in several disciplines of science and engineering. The performance of a data grid system for real-time applications is highly dependent on the underlying job scheduling, data scheduling, and data replication algorithms and advance reservation mechanism. Thus, in the literature, there are numerous studies that propose solutions to the job/data scheduling, data replication, and advance reservation problems. In these studies, a number of simulators, emulators, or test beds have been used to evaluate the proposed algorithms. Furthermore, these simulators/emulators usually adopt fixed-grid models, which in turn dictate specific job/data scheduling and data replication mechanisms. In the literature, there is no unified framework for modeling grid systems with different architectures, which can allow researchers to develop new grid system models and evaluate them in a flexible manner. This paper presents a unique framework for modeling real-time data grid systems that attempts to unify a large class of job scheduling, data scheduling, and data replication algorithms based on several system services. Then, in order to enable the development of these algorithms under different system models, DGridSim is realized to be a multi-model discrete-event simulator, and its capabilities are exemplified by means of a set of simulation results. The main contribution of the research is DGridSim, which can model and simulate a variety of different data grid system models by means of several system services and their interactions.