Abstract: Traditional architectures of decision support system are most based on database cluster of SMP or MPP with a high cost of storage and computing. It is difficult to expand, and the performance bottleneck of large scale offline computing and streaming computing. By analyzing the challenges of DSS, this paper puts forward a fusion architecture of DSS based on big data, which satisfies the requirements of relational operations, large-scale offline distributed computing and streaming computing. It supports the analysis and processing of structured data, semi-structured data and unstructured data.