Parallelisation

Calibration requires many model runs.

PEST is able to distribute these runs across many computing nodes, and/or across many different computers. If desired, these computers can be dedicated solely to the running of models under the control of PEST. Alternatively, they can be ordinary personal computers spread throughout a normal office to which PEST is given temporary access during the calibration process.

By spreading model runs across an office network, or the nodes of a LINUX cluster, the time required for calibration of a model can be reduced enormously.

Where run parallelisation is used in conjunction with PEST’s unique “SVD-Assist” inversion engine, state-of-the-art calibration and uncertainty analysis methods can be applied even to complex models with long run times. Without PEST, this would not be possible.