Abstract. In the future, advanced biomechanical simulations of brain deformation during surgery will require access to multi-teraflop parallel hardware, supporting operating room infrastructure. This will allow surgeons to view images of intra-operative brain deformation within the strict time constraints of the surgical procedure - typically on the order of minutes, multiple times during a six or eight hour long surgery. In this paper we explore the grid infrastructure issues involved in scheduling, on-demand computing, data transfer and parallel finite element biomechanical simulation, which would guarantee that such a dynamic data driven real time application is actually feasible.