This document provides a summary of key concepts related to data sharding and replication in YugabyteDB. It discusses three methods for sharding data across tablets: pre-splitting tablets at table creation, manual splitting of tablets at runtime, and automatic splitting of tablets as they grow beyond a size threshold. It also explains how YugabyteDB uses Raft consensus for replication, with tablets replicated across multiple nodes to provide fault tolerance. Transaction processing is achieved by writing provisional records to leader tablets and committing transactions once a majority is reached.