My every process needs to read RocksDB, causing high memory usage when multiple processes are launched on a single machine. Can a shared memory approach be implemented?

[ljfgrace]

My use case is read-only. RocksDB occupies 19 GB, and each process that loads it into memory consumes approximately 1 GB, with no available means to reduce per-process memory usage. Is there an official mechanism provided for inter-process memory sharing?