Case Study: GumGum achieves scalable, lower-cost, low-latency ad targeting with Amazon DynamoDB

GumGum - Customer Case Study

GumGum, a high‑traffic programmatic advertising platform, faced growing operational pain running Apache Cassandra on 106 i3.2xlarge nodes: manual scaling, data‑center outages, increasing engineering fatigue and revenue risk from any latency spikes (90% of traffic comes from programmatic partners, so low response time was critical).

GumGum migrated to Amazon DynamoDB — using parallel writes (write→wait→read) for behavioral data, EMR + S3 + Spark for contextual data, DAX/Memcached caching, and master‑master/Global Tables replication across regions. The move delivered sustained low latency (<3–5 ms), handled ~8 million reads/min across 16.2 billion items (2 TB), produced zero outages and fewer timeouts, eliminated throttling, and reduced annual infrastructure costs by roughly 60–65% (from ~$437K to ~$162K).

GumGum

Anirban Roy

Lead Engineer