Performance at Scale: Benchmarking FHIR Servers Under Real-World Load
FHIRサーバーの性能評価：実環境負荷でのベンチマーク結果を公開
Summary
This article outlines a comprehensive benchmarking methodology for FHIR servers, testing performance across three stages: empty database, 100K data load, and incremental growth. It validates scalability and operational cost impact by providing practical performance metrics for core workloads (CRUD, batch processing, search).
Details
Performance is critical when selecting a FHIR server, impacting not only user experience but also infrastructure costs like database size and compute resources. This benchmark compares multiple open-source FHIR servers against Aidbox, measuring key metrics such as throughput (operations per second), latency (p99 response times), and resource consumption. The test suite is structured in three stages. Stage 1 establishes a baseline from an empty database by running CRUD operations and batch imports. Stage 2 simulates mid-size deployment by loading 100,000 synthetic patient records, assessing sustained write pressure handling. Crucially, Stage 3 measures how query performance changes when data is incrementally added to the existing dataset (the delta between stages), revealing the server's true ability to maintain performance as data volume grows in a real-world setting.
Original content copyright by respective publishers