The meteorological data is widely varied, massive, and complex, so it has made huge system pressure in the China Integrated Meteorological Information Service System (CIMISS). The design of database tables and the optimization of the process are analyzed in detail by the Lband upperair data as a case. The database redundancy is reduced by means of the "keyelement," reasonable index, and partition to improve database access efficiency, through simple procedure to reduce the message interaction between the system and IO, so as to optimize the storage efficiency. Comparison of the storage methods OCI and PRO*C shows that PRO*C is easy to develop, and OCI has comprehensive functions but relatively complex. By comparing the efficiency of two storage modes, the system takes a PRO*C way. The actual running effect of Lband data shows that to complete a data processing takes an average of 28 s. Compared with the sounding data, the simple procedure performs faster, compared to the original.