Abstract: A unified storage and retrieval method was developed for space weather scientific data of the Chinese Meridian Project to address massive small-file management and multi-granularity access requirements. Object storage was used as the underlying platform, and columnar containers were adopted as the basic data organization unit. Discrete files were grouped by data product category and temporal partition, and original file contents were stored together with parsed structured records derived from the same file within each container. An external metadata index was combined with statistical pruning inside containers to support both file-level trace-back retrieval and record-level conditional query while reducing unnecessary data scanning. The method was validated using real observational data collected over four years, with a total volume of approximately 2.4 TB and about 15.1 billion records. Results showed that the overall compression ratio reached 29.9%, and text data achieved a compression ratio of 46.2%. For parseable numerical data, about 2 million original files were consolidated into 93 containers, and record density per unit storage space increased by about 80%. Under concurrent access conditions, both services remained stable. The method alleviates the pressure of object and metadata management and provides a reference for the unified organization and online service of similar scientific observation data.