Evaluating a $2M Commercial Server on a $2K PC and Related Challenges

Talk Abstract Many future multiprocessor servers will execute large commercial workloads, such as database management systems and web servers. Thus, simulations of new multiprocessor designs should run these workloads. However, simulating expensive servers running these large workloads on low-cost personal computers presents many challenges. Most of this talk discusses the Wisconsin Multifacet project's approach to commercial workload simulation [1]: Multifacet's Simulation Infrastructure and Workload Development Process. Our infrastructure uses real hardware , locally-developed simulators, and Virtutech Simics. Our workload development process selects commercial workloads, tunes them on real hardware, scales them for simulation, and then re-tunes the scaled workloads. Managing Simulator Complexity. We separate the timing simulator from the functional simulator. The timing sim-ulator mimics the micro-architectural timing of alternative future systems, along with the functional behavior of dynamically common instructions. The functional simulator, Simics, checks the timing simulator by modeling full-system functionality. Coping with Workload Variability. Small changes in micro-architectural parameters can cause multithreaded work-loads to follow different paths (e.g., due to lock race and OS scheduling). We add pseudo-randomness to our deter-ministic simulation to enable standard statistical analyses to distinguish significant effects from random ones. that include making simulators more modular, developing benchmarks more appropriate to target areas, expanding the use of analytic models to complement simulation, and increased emphasis on validating results via multiple methods. The results I present reflect the creativity of the authors of the original works, whereas errors in this talk are mine.