Skip to content

SYSTEM Cited by 1 source

Uber Fulfillment Platform

The Fulfillment Platform is Uber's 2021+ ground-up rewrite of its dispatch + fulfillment stack on a NewSQL architecture using Google Cloud Spanner as the primary storage engine. Uber's first major public bet on a cloud-native primary storage layer for a load-bearing OLTP workload. (Source: sources/2024-03-14-highscalability-brief-history-of-scaling-uber)

Why rewrite

By 2021, the original dispatch stack couldn't gracefully support the full range of Uber use cases:

  • Reservation flows — driver confirmed upfront.
  • Batching flows — multiple trips offered simultaneously.
  • Virtual queue mechanisms at airports.
  • Three-sided marketplace for Uber Eats (consumer × restaurant × driver).
  • Package delivery through Uber Direct.

Each was bolted onto assumptions the original matching engine made for simple rider-driver pairings. A clean rebuild was needed.

Why Spanner

Lead Uber engineer Ankit Srivastava's framing (quoted in the 2024-03-14 retrospective):

"as we scale and expand our global footprint, Spanner's scalability & low operational cost is invaluable. Prior to integrating Spanner, our data management framework demanded a lot of oversight and operational effort, escalating both complexity and expenditure."

The pick targets transactional consistency, horizontal scalability, and low operational overhead as the core tradeoff triangle — classic NewSQL/Spanner fit.

Primary-source reference: Building Uber's Fulfillment Platform (with Spanner).

Seen in

Last updated · 517 distilled / 1,221 read