This study's objective is to solve mining knowledge issue from different sources of distributed, formatted or unformatted data with diverse data semantics. The issue is formulated as data mining in knowledge grid. In particular, this paper presented a DM-GRID (data mining in grid) model based on software architecture of a novel infrastructure for distributed and high-performance data mining in knowledge grid. The DM-GRID model is characterized by transparent conversions between heterogeneous data formats using MDI (metadata information), communication models from one to the other sources of data using broadcasting, and united programming interfaces. This model of Data Mining GRID conducts an access- processing application on grids. To study the effects of the proposed model, the experiment based on mining association rules in PC cluster are designed and implemented. The result shows that DB-GRID is feasible. This study's conclusions indicate that model could provide a possible thoughtful method of integration on large-scale distributed data mining.