Playback: Real-time targeting with MongoDB at Braze - MongoDB World 2018

Dj Walker-Morgan


There were a lot of great talks at MongoDB World 2018 from people running MongoDB in production and at scale. We'll be running through some selected talks here, in what we call Playback.

Braze takes on the challenge of processing billions of events into targeted messages to consumers and powers their business with MongoDB. It's critical to do this at speed, as Braze CTO and Co-founder Jon Hyman explains in this MongoDB World 2018 presentation, as data's value starts reducing as soon as you've captured it.

He dives into the MongoDB architecture and data models that let Braze manage almost 10 billion user document, with a 1,000 shards and 50 different clusters across the US. Storage isn't the issue for Braze though, it's throughput and across their clusters, they run at 263,000 MongoDB operations per second. Learn how Braze deploys MongoDB so that it fits each of their clients' profiles, from shared multi-tenancy for smaller users up to large-scale dedicated clusters.

In what may surprise some developers, Braze tends not to have indexes for customer queries. Hyman talks about how their implementation takes on handling queries efficiently without indices by bucketing users and parallelizing queries to match the demands their clients have. He also looks at how they use single documents to retain client metadata and campaigns. It's a great look at how operation intensive workloads are handled with MongoDB.