Hippocampus: A Fast-Scalable Database Engine

This paper describes the design and implementation of an in-memory database engine built in Golang, prioritizing performance and data durability. The system supports string and hash data structures, with client-server communication handled via a Hippocampus Serialization Protocol (HISP) parser. HISP’s binary-safe design, using length-prefixed bulk data transfers, ensures efficient serialization of integers, strings, and arrays while supporting cross-platform compatibility. For data persistence, the engine employs an Append-Only File (AOF) mechanism to log transactions continuously, enabling reliable recovery after system failures. The paper examines three deployment approaches—local, Docker-based, and Kubernetes-based—to meet varying scalability needs. The proposed engine is compared with existing in-memory databases, including GorillaDB, Memcached (used by Facebook), and FoundationDB (adopted by Apple, Snowflake, and VMware). The analysis evaluates performance, scalability, and features to situate the engine within the broader context of scalable database systems.