← All runs

Round 0

2026-04-08

Apple M3, 8-core · 24 GB · macOS Tahoe 26.3 · Docker Desktop 4.64.0 (8 CPU / 20 GB)

SNOMED combined filters

POST /ValueSet/$expand with composed ValueSet

The most complex expansion test. Each request combines three filter types simultaneously: an is-a hierarchy constraint, a SNOMED property equals filter (e.g. associated morphology), and a free-text display filter. Tests the server’s ability to efficiently intersect multiple constraints at once.

Preflight request
POST /ValueSet/$expand
Content-Type: application/fhir+json
Accept: application/fhir+json

{
  "resourceType": "Parameters",
  "parameter": [
    {
      "name": "valueSet",
      "resource": {
        "resourceType": "ValueSet",
        "compose": {
          "include": [
            {
              "system": "http://snomed.info/sct",
              "filter": [
                { "property": "concept",   "op": "is-a", "value": "404684003" },
                { "property": "116676008", "op": "=",    "value": "72704001"  }
              ]
            }
          ]
        }
      }
    },
    { "name": "filter", "valueString": "fracture" },
    { "name": "count",  "valueInteger": 10 }
  ]
}
Ranking
Servers ranked by weighted RPS (wRPS) for this test. wRPS = raw RPS × per-test normalizer × bias factor. Score % is relative to the top performer. Raw RPS is the measured peak throughput; imputed values (shown in italics in orange) are estimated for servers that did not run this test. Status: ✓ preflight passed, ✗ failed, ~ skipped.
# Server wRPS Score RPS Errors Status
1 termbox 7.9K
100%
1.1K 0%
2 ontoserver 45
1%
6 0%
3 fhirsmith 24
0%
3 0%
4 snowstorm 11
0%
1
5 hades 11
0%
1 ~
Latency
Latency percentiles per server across all VU levels. Columns: percentile / VU count.
p50/1 p95/1 p99/1 p50/10 p95/10 p99/10 p50/50 p95/50 p99/50
termbox 3.3ms5.5ms7.7ms8.8ms12ms15ms45ms51ms55ms
fhirsmith 350ms403ms468ms2.76s5.59s11.07s3.88s41.07s43.11s
ontoserver 504ms1.07s2.21s1.55s3.24s3.64s11.33s13.84s15.00s