What database was Discord originally built on?

Discord was originally built on MongoDB. The first version of Discord was created in early 2015, and its early data was stored in a single MongoDB replica set.

MongoDB suited the fast-moving startup phase because it is a document-oriented NoSQL database, allowing Discord’s developers to iterate quickly without designing a large relational schema. As usage expanded, however, Discord’s message collection and indexes became too large to keep comfortably in memory, and latency became less predictable.

Discord then moved message storage to Apache Cassandra, a distributed wide-column database designed for large-scale, fault-tolerant workloads. Later, the company migrated its databases to ScyllaDB, a Cassandra-compatible system written in C++. The progression is therefore MongoDB first, Cassandra next, and ScyllaDB later—not MongoDB as Discord’s permanent database.

A common mix-up is confusing Discord’s current infrastructure with its original technology. Discord now uses several data systems, including ScyllaDB, PostgreSQL, and Elasticsearch, depending on the workload.