Redpanda employs the Raft distributed consensus algorithm to manage its distributed log system, initially starting with a basic implementation that lacked features like reconfiguration but has since been extended to improve scalability, reliability, and speed. Redpanda supports Kafka's ACKS property, allowing different consistency and durability levels, and initially used varied approaches for handling these levels. However, the need for a simpler, unified Raft replication mechanism led to the development of a solution that treats all ACKS values uniformly using replicate_batcher and replicate_entries_stm, simplifying code, enhancing maintainability, and improving performance by controlling backpressure and memory usage. This new approach optimizes replication latency, reduces storage pressure, and allows parallel handling of requests, thereby facilitating future enhancements like in-memory writes and remote leaders. By unifying the code for Raft replication, Redpanda ensures that all optimizations apply across all consistency levels, improving overall performance while maintaining flexibility for future developments.