Detailed simulation of many discrete event systems is often slow even on state-of-the-art workstations. Parallel discrete event simulation (PDES) techniques offer an alternative to speed up such simulations. However, PDES is a hard problem. The irregular, data-dependent nature of most simulations make synchronization a difficult issue. In this CAREER project an inter-disciplinary approach will be taken to address this problem. A new class of "adaptive" protocols will be developed that integrates several mechanisms in a single framework, to achieve optimal performance automatically, notwithstanding the dynamic nature of most simulations. The target platform is a network of workstations or possibly multiprocessors, connected by a high-speed network. Several innovations are planned regarding protocol development and implementation in a PDES kernel on such platforms. These include accounting for the multi-granular of the platform and use of fast, user-level messaging support. The work will focus on simulating of telecommunication and computer networks.
