HL7 FHIR server selection involves specific criteria. Four cover most decision space.
1. Bulk export sustained throughput. Bulk Data IG $export at 10M+ resources. HAPI, Aidbox, Medplum all scale; smaller vendors hit walls.
2. Terminology co-location. In-process (HAPI, Aidbox) simpler; standalone (Ontoserver) more capable.
3. Subscription reliability. Subscription retry, dead-letter, back-pressure.
4. Observability. Prometheus per resource type.
Vendor decision matrix
| Server | Bulk export | Terminology | Subscription | Metrics |
|---|---|---|---|---|
| HAPI JPA 7.x | Good with tuning | In-process | External | JMX |
| Aidbox 2409 | Excellent | In-process | Built-in | Prometheus |
| Medplum 3.x | Excellent | Minimal | Built-in | OpenTelemetry |
| Microsoft FHIR | Fair | External | Add-on | Azure Monitor |
| InterSystems IRIS | Excellent | Enterprise | Enterprise | Enterprise |
Common selection mistakes
1. Peak throughput over sustained. 2. Ignoring terminology. 3. Underestimating auth. 4. Missing observability. 5. Assuming vendor claims.
Investment sizing
1. HAPI (open): dev + ops. 2. Aidbox: $200-400k/year. 3. Medplum: $150-300k/year. 4. Firely: $150-300k/year.
HL7 FHIR server selection is a 5-year commitment. Verify with load tests + Inferno + reference customers before commit.
