Design an ANN index system

Design a system that can efficiently retrieve the IDs of the most similar ~50 vectors to a given query vector from a dataset of ~10 billion vectors, supporting high-throughput read and write operations with approximate nearest neighbor search capabilities.