Why does Gurobi perform differently on different machines?

Solving the same model on different machines can result in different solution paths.

This effect is a well-known phenomenon called "Performance Variability". Hardware plays an important role as the search path within the branch-and-bound tree can be significantly different. In addition, changing the order of variables or constraints or using random seeds can also drastically impact the performance even when the mathematical models are identical. This is described in theMIPLIB 2010 paperby Koch et al.

The reasons for performance variability are quite diverse. Gurobi makes many decisions in the search for an optimal solution. Whenever there is a decision that is a tie or even close to a tie, a difference in the computing environment, even if it is seemingly insignificant, can lead to a different outcome. These differences can originate from differences in operating systems, underlying libraries, computer hardware, etc.

Gurobi attempts to exploit this when running inConcurrentMIP mode.

To get a sense of how susceptible your model is in regards to performance variability, you can compare the performance of several runs that only differ in therandom seed parameter.

Further information

• Is Gurobi Optimizer deterministic?
• What hardware should I select when running Gurobi?