SYSTEM Cited by 1 source

Schemaless (Uber)

Schemaless is Uber's trip data store, built in 2014 to replace a single-instance PostgreSQL that was becoming an existential bottleneck — the majority of the DB was trip data, growing fastest, and six months from the 2014 Halloween peak. Schemaless is "an append-only sparse three-dimensional persistent hash map, similar to Google's Bigtable, built on top of MySQL" — horizontal-sharded by row for rapid development and scale. (Source: sources/2024-03-14-highscalability-brief-history-of-scaling-uber)

The migration succeeded in averting the Halloween peak disaster — now a piece of Uber engineering lore. Canonical case study of "the right primitive is often a well-known concept rebuilt on boringly-reliable infrastructure" (Bigtable shape + MySQL storage).

Blog lineage (primary-source links from Clemm's retrospective):

• Schemaless design — part one: MySQL datastore
• Postgres to MySQL migration
• Mezzanine codebase data migration — the Halloween-deadline migration playbook.
• Rene Schmidt video on Schemaless creation + migration

Seen in

• sources/2024-03-14-highscalability-brief-history-of-scaling-uber — introduced as the 2014 trip-data-store rewrite; migrated all trip-touching services to Schemaless-on-MySQL in time for Halloween. Other non-trip workloads moved to Cassandra.

Related

• systems/mysql — storage substrate.
• systems/google-bigtable — structural model (sparse 3D hash map).
• systems/apache-cassandra — other 2014 data-store bet at Uber; used for marketplace matching/dispatching.
• concepts/horizontal-sharding — row-level horizontal partitioning.
• companies/uber — origin org.