Abstract:The DM data in the large scale and complex dimensions, to as much as possible in limited conditions to meet the needs of users of the DM database function, proposes a scalable parallel algorithm of large-scale new data in the DM database. Not a scalable parallel algorithm including simple parallel, parallel and parallel three kinds of typical logic processing rules, a new algorithm of the three kinds of rules to combine data independent operations, so that each computation node has three processing modes, using directed graph divide the large-scale data into local data, and assigned to the processor, through set the priority of data processing, to complete the pipeline in the form of data processing, with strong scalability of parallel algorithms. The experimental results show that the new algorithm has strong scalability and excellent debt balance ability.