Currently, jobs are limited to walltime of 72 hours (phases 7 and up exclude phase 8c) and 168 hours (phases 1-6 and 8c). If you expect that your computations may take more time, you have several options:

  • Checkpoint your application: this approach is often useful in large/long-running simulations. Often it is possible to periodically save intermediate results from the application, so that after a job terminates, the application can be “restarted” from the latest saved results.

  • Optimize your application: it is often possible to obtain better performance by various means, such as parallelizing serial programs, using GPUs, optimizing I/O, etc., The first step in such optimization is always to identify bottlenecks, i.e., identifying which portion(s) of the application/program is/are the largest contributor to the running time. You are welcome to visit the Office Hours or schedule a meeting with our research support staff to discuss optimization

  • Submit a reservation request