Some of you will have seen this, if you haven't it's well worth a read to see what's cooking in DOLFIN with regards to parallel scaling: http://www.hector.ac.uk/cse/distributedcse/reports/UniDOLFIN/UniDOLFIN.pdf It addresses the lack of (i) scalable parallel I/O, (ii) parallel mesh refinement and (iii) hybrid MPI + OpenMP in DOLFIN. Florian