The distributed constraint optimization problem (DCOP) is able to model a wide variety of distributed reasoning problems that arise in multiagent systems (MAS), and distributed algorithms of solving DCOP have already become one of the most important bases of MAS. Some previous algorithms, which emphasize the asynchronous communication, distributed computation and quality guarantees, such as Adopt, can obtain optimal solution of DCOP by negotiation among peer to peer agents. However, there is still possibility to improve in organization structure of solving problems. A novel organization structure of multi-agent is put forward, which adopts the idea of combination of decentralization and centralization, the partitions of constraint graph method and the notions of core node, and the main communication road. The asynchronous and distributed algorithm of DCOP in this organization structure can improve the efficiency in execution and adaptation in dynamics. Moreover, it can unite the solving method of DCOP dealing with a variable per agent and multiple variables per agent.