The purpose of mathematical reservoir simulation models in petroleum applications is to try to optimize the recovery of hydrocarbon from permeable underground reservoirs. To accomplish this, one must be able to predict the performance of the reservoir under various production schemes. There are two essential issues, modeling and software architecture design, while developing a comprehensive oil reservoir modeling platform that should be an integration of subsurface models, facility network models and economic models. Effective subsurface models must be constructed to describe the complex geomechanical, physical, and multiphase fluid flow processes that accompany the various recovery mechanisms. Upscaling needs to be utilized to provide effective rock properties for coarse-grid models used for field-scale simulations. However, localized flow regimes at sub-coarse grid scales must often be resolved using local grid refinement techniques. Finite volume element methods for accurate resolution of localized geometrics can be coupled with cell-centered finite difference methods used in many existing simulators. Aspects of coupling different grids, different discretization schemes, and different physical equations via mortar techniques will be presented. Reservoir simulation is an integration of various technologies through the construction of a reservoir model as well as optimization of production strategies. A comprehensive oil reservoir modeling platform should be an integration of different software applications or components and its software architecture should be scaleable, extendable and should have the capability to create and modify a workflow. Beyond the traditional three-tier software architecture, data, application, and user-interface, separation of control and business logic through those three tiers is proposed to achieve those goals. The aspect of the software architecture design will be discussed.