Microsoft Azure Health Data Services · Rate Limits

Microsoft Azure Health Data Services Rate Limits

Azure Health Data Services enforces rate limits per service. FHIR service throughput scales with provisioned RU/s (automatically managed by Azure — not manually configurable). DICOM service handles thousands of image ingestions per second but doesn't publish explicit per-second request caps. The de-identification service throttles at 1 MB per 5 seconds or 100 requests per 5 seconds. When limits are exceeded, the service returns HTTP 429 (Too Many Requests). Higher throughput is available by submitting a support request.

Microsoft Azure Health Data Services Rate Limits is the machine-readable rate-limit profile for Microsoft Azure Health Data Services on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 9 rate-limit definitions, measuring payload_size, mb_per_window, requests_per_window, documents, and mb.

The profile also includes 3 backoff/retry policies defined and response codes documented for rateLimitExceeded and serviceUnavailable.

Tagged areas include Rate Limiting, FHIR, DICOM, De-identification, and Healthcare.

9 Limits
Rate LimitingFHIRDICOMDe-identificationHealthcareMicrosoft Azure

Limits

De-identification service — request size request
payload_size · per-request
50 KB
Each synchronous de-identification request body must not exceed 50 KB.
De-identification service — throughput (bandwidth) service-instance
mb_per_window · 5_seconds
1
Requests are throttled when total data volume exceeds 1 MB per 5-second window. Submit a support request for higher throughput.
De-identification service — request rate service-instance
requests_per_window · 5_seconds
100
Requests are throttled when request count exceeds 100 per 5-second window. Submit a support request for higher limits.
De-identification service — async job documents job
documents · per-job
10000
Each asynchronous batch job can process no more than 10,000 documents.
De-identification service — document size (async jobs) document
payload_size · per-document
2 MB
Each individual document submitted in an async batch job must not exceed 2 MB.
De-identification service — free monthly allotment service-instance
mb · month
50
50 MB of de-identification included free each month per service instance before usage-based charges apply.
FHIR service — throughput (RU/s) service-instance
request_units_per_second · second
-1
Throughput is automatically managed by Azure and scales elastically. Manual RU/s provisioning is not available in Azure Health Data Services FHIR service (unlike Azure API for FHIR). Contact support for throughput capacity discussions.
DICOM service — ingestion capacity service-instance
images_per_second · second
-1
Designed to ingest multiple thousands of DICOM images per second. Exact cap is not published; scales to petabyte-level imaging archives. Contact support for capacity planning.
MedTech service — device messages service-instance
messages_per_second · second
-1
No documented hard cap per second; throughput scales with the Azure Event Hub partition count configured as the device message input.

Policies

Retry on 429 with exponential backoff
When the de-identification service returns HTTP 429, clients should implement exponential backoff with jitter before retrying. The Retry-After header may indicate the recommended wait time.
Support request for higher throughput
For FHIR, DICOM, or de-identification workloads requiring throughput above default limits, submit an Azure support request for consideration of higher capacity allocations.
Private Link for network isolation
Enable Azure Private Link to route all API traffic through private endpoints, reducing exposure to public internet throttling and network-layer limitations.

Sources