The estimated cost equation seems to assumes that the area of the model domain will scale as the number of state vector elements. The number of state vector elements dictates the number of simulations to be done and the area of the model domain, both of which are linear in cost. But the model domain also includes the area occupied by the buffer elements, which doesn’t scale with the number of state vector elements, and at least for the Permian most of the model domain is dedicated to these buffer elements. So the equation should be corrected to account for the fraction of the model domain occupied by the buffer elements in the reference simulation.
Currently, the IMI uses the following method to calculate the estimated cost on AWS:
The estimated cost equation seems to assumes that the area of the model domain will scale as the number of state vector elements. The number of state vector elements dictates the number of simulations to be done and the area of the model domain, both of which are linear in cost. But the model domain also includes the area occupied by the buffer elements, which doesn’t scale with the number of state vector elements, and at least for the Permian most of the model domain is dedicated to these buffer elements. So the equation should be corrected to account for the fraction of the model domain occupied by the buffer elements in the reference simulation.