Design of LAMOST data processing and production database

Before LAMOST spectra release, raw data need to go through a series of processes, i.e. a pipeline after observed, including 2D reduction, spectral analysis, eyeball identification. It is a proper strategy that utilizing a database to integrate them. By using database the coupling between relative modules would be reduced to make the adding or removing of them more convenient, and the dataflow seems to be more clearly. The information of a specific object, from target selection to intermediate results and spectrum production, can be efficiently accessed and traced back through the database search, rather than via FITS reading. Furthermore, since the pipeline has not been perfected yet, the eyeball check is needed before the spectra are released, and an appropriate database can make the feedback period of eyeball check result more conveniently, thus the improvement of the pipeline will be more purposely. Finally, database can be a data mining tools for the statistics and analysis of massive astronomical data. This article focuses on the database design and the data processing flow built on it for LAMOST. The database design requirement of the existing routines, such as input/output, the relationship or dependence between them is introduced. Accordingly, the database structure suited for multiple version data process and eyeball verification is presented. The dataflow, how the pipeline is integrated relied on such a dedicated database system and how it worked are also explained. In addition, some user interfaces, eyeball check interfaces, statistical functions are also presented.