The adaptive mesh refinement is one of the main techniques used for the solution of partial differential equations. Since 3-dimensional structures are more complex, there are few refinement methods especially for parallel environments. On the other hand, many algorithms have been proposed for 2-dimensional structures. We analyzed the Rivara's longest-edge bisection algorithm, studied parallelization techniques for the problem, and presented a parallel methodology for the refinement of non-uniform tetrahedral meshes. The main goal of this research is to propose a practical refinement framework for real-life applications. We describe a usable data structure for distributed environments and present a utility capable of distributing the mesh data among processors to solve large mesh structures.