One of the important problems in supply chain management is the coordination of the inventory control of each stage in a supply chain. Presented was an approach to optimization of a 2-echelon inventory system in a supply chain, including Markov decision process(MDP), Markov games(MG) and an intensive learning algorithm to solve the MG problem. In particular, the 2-echclon system was modeled as a MG and the concepts of Markov Games and intensive learning algorithm were introduced to give the optimal solution of the inventory system.