As of the release of Distributed Computing Toolbox 2.0 and the MATLAB Distributed Computing Engine 2.0, both independent and interdependent tasks can be executed. However, for interdependent tasks, you must use either the mpiexec scheduler or the MathWorks job manager.
In previous releases, the Distributed Computing Toolbox and MATLAB Distributed Computing Engine supported only coarse-grained applications in which tasks are executed independently of each other on the cluster, i.e., "embarrassingly parallel" applications.
To learn more about the two products, see the following Web pages:
Best Answer