Abstract: This paper proposes a low management cost computing grid model，aims at harvesting the idle time of Internet connected computers to run very large distributed applications.In this grid model，no specific node is designated to manage the dynamically changing resources.In the absence of the management node，the corresponding information mechanism，task schedule algorithm and limited task replication algorithm naturally evolve the system into equilibrium to perform large scale distributed computation with low cost.A simulation software package has been developed to verify this model，and to assess its performance.The results show that the model reasonably distributes loads in a dynamic environment，and fully utilizes the computational capacity of the idling resources，it is a high-efficiency resolution for large scale，embarrassingly parallel applications such as parameter sweep and Monte-Carlo simulations.