Distributed Computing

Distributed computing enables models with multiple cases to be divided among the many cores of a cluster with multiple execution nodes, resulting in significant reduction in overall time to obtain results. The run time scales nearly linearly with the number of cores, making this especially valuable for DOE and optimization runs.

It also provides a convenient way to efficiently use all available GT-SUITE licenses, even for fast running simulations with smaller case counts. Users submit jobs to the job scheduler, which splits the jobs into smaller “packets”. The scheduler intelligently queues packets until licenses become available and then submits them to the execution nodes, reducing license idle time. After all the packets of a given simulation are complete the results are automatically combined into a single file, just as if the model was run locally.

The distributed computing software is provided at no additional charge. Each packet running on an execution node requires only a solver license (no GUI license). Consequently, reduced cost “solver-only” licenses are available. It is very common for customers to have significantly more solver licenses than GUI licenses.

The same technology is also available for single machines with multiple cores. Users can quickly and easily run multi-case jobs on multiple cores on their desktop machine.

Note that, in addition to distributed computing, where multi-case runs are divided in to packets of one or more cases, GT-SUITE offers a parallel solver option which allows a single case to be run in parallel on multiple cores.


Contact Us

Ready to try GT-SUITE?