much better solution is to build a brand-new database inside the batch job and write it as files to the job’s output directory in the distributed filesystem, just like the search indexes in the last section. Those data files are then immutable once written, and can be loaded in bulk into servers that handle read-only queries.
