This paper proposes a new heuristic method for the logistics network design and planning problem based on linear relaxation and DC (difference of convex functions) programming. We consider a multi-period, multi-echelon, multi-commodity and multi-product problem defined as a large scale mixed integer linear programming (MILP) model. The method is experimented on data sets of various size. The numerical results validate the efficiency of the heuristic for instances with up to several dozens facilities, 18 products and 270 retailers.