How a team at Epic Games tuned Amazon OpenSearch Service for Fortnite analytics

August 10, 2026 - 22:58

Behind the scenes of Fortnite, a massive amount of gameplay data flows through Epic's analytics systems every second. To keep that pipeline fast and cost-effective, the team recently overhauled how they run Amazon OpenSearch Service. Working alongside AWS specialists, they focused on a few key areas that made a big difference.

The first step was right-sizing the instances. The old setup used a mix of compute and memory that did not match the actual workload. By testing different instance families and picking ones that fit the query patterns better, they cut waste without losing performance. Next came shard rebalancing. Some indexes had too many small shards, which created overhead. Others had too few, causing hot spots. The team redistributed shards so each node carried a more even load, which smoothed out latency spikes during peak play times.

Index mappings also got a closer look. Many fields were indexed even though they were rarely queried. By disabling indexing for those fields and using more efficient data types, they reduced storage pressure and sped up writes. That change alone trimmed the cluster's memory footprint.

The bigger win came from upgrading the engine version and moving to Graviton-based instances. The newer OpenSearch engine included better query planner logic, and Graviton processors offered a better price-to-performance ratio for this kind of throughput. After the move, query latency dropped noticeably, and the cluster handled more concurrent searches without throttling.

The end result was a cluster that responds faster, scales more predictably, and costs less to run. For a game with millions of active players, that means analytics dashboards stay fresh, and the backend team can spend less time firefighting and more time building features.