Case Study: mParticle achieves real-time processing and eliminates backlogs with ScyllaDB

mParticle Improves System Performance by Migrating to Scylla

mParticle, a customer data platform that ingests over 50 billion messages, 100 billion events and 150 TB of raw data per month for customers like NBC and Spotify, faced severe performance and SLA issues with their Cassandra-based user profile store and rules engine. Cassandra’s pending compactions during batch loads caused high read/write latency, processing backlogs, complex operational tuning, and costly support gaps, so mParticle evaluated alternatives and turned to ScyllaDB to address these challenges.

mParticle ran a POC and migrated clients to ScyllaDB one at a time, finding no code or data-model changes were needed and that ScyllaDB dramatically outperformed Cassandra for their workload. ScyllaDB’s self-tuning, compaction isolation, and strong engineering support eliminated backlogs—work that took 20 hours on Cassandra now streams in real time—allowing mParticle to meet SLAs with far less operational effort.

mParticle

Nayden Kolev

Systems Architect