**Task:**

- solve systems of partial differential equations with nonlinear coupling
- background: chemical process engineering
- problem with parameter studies: a few modes become unstable
- fast standard solvers don't work
- other solvers work slowly

**Basic Idea:**

- split the solution space into a (small) hard part and a (huge) easy part
- use different (parallel) solvers in each subspace
- synchronize both solvers when necessary

**Difficult load balancing:**

- solvers have different convergence speed
- synchronization using traditional messages difficult (e.g. via polling)

**Solution:**

- solvers write results directly into control task
- use onesided communication