Forum Discussion
API data sources
Hi Experts, we have to ingest the data from API , and we have options like direct API mode or APIM . There are nearly 10 data sources, and APIM would cost us. So which options we should go for on which scenario, do we have any benchmark criteria in terms of performance, and costs.
Hi SamyAbdul,
v-saisrao-msft' reply is the approach I would take as well.
I would not use the number of APIs alone as the deciding factor. Ten APIs can still be perfectly manageable with direct Fabric REST ingestion if Fabric is the main consumer and the APIs have straightforward authentication, pagination and rate limits.
Microsoft’s Fabric REST connector already supports pipeline Copy activity with pagination and authentication options including service principal.
I would bring APIM in when you need an API management layer rather than just an ingestion mechanism - for example centralized authentication, throttling, quotas, policy enforcement, transformations, monitoring, or multiple consumers sharing the same APIs.
For cost/performance benchmarking, I would compare:
- request volume and concurrency;
- payload size and pagination;
- API rate limits;
- retry/error rates;
- end-to-end ingestion duration;
- APIM tier/capacity required;
- number of consumers using the APIs.
Microsoft’s API Management cost guidance is useful here because APIM cost depends on the tier and capacity/request model rather than simply the number of APIs.I would therefore start direct unless there is a clear governance or shared-platform requirement, then introduce APIM when the API layer itself needs centralized management.
AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.
2 Replies
- v-saisrao-msft
Community Support
Hi SamyAbdul,
For your scenario with around 10 API sources, I recommend starting with direct API ingestion using Microsoft Fabric Data Factory's REST Copy activity, provided the APIs meet the connector's authentication and connectivity requirements. Configure pagination, request intervals, timeouts, and retries as appropriate for each API. Consider APIM when you require centralized authentication, rate limiting, API governance, or shared policies across multiple consumers. Before deciding, compare the total cost of APIM against direct ingestion and benchmark both approaches using representative APIs, measuring ingestion duration, latency, error rates, and request volume. If direct ingestion meets your security, performance, and reliability requirements, APIM may not be necessary.
REST connector overview - Microsoft Fabric | Microsoft Learn
Configure REST in a copy activity - Microsoft Fabric | Microsoft Learn
Plan and Manage Costs for Azure API Management | Microsoft Learn
Azure API Management Policy Reference | Microsoft Learn
Thank you. - ShivekMaharaj
Community Champion
Hi SamyAbdul,
v-saisrao-msft' reply is the approach I would take as well.
I would not use the number of APIs alone as the deciding factor. Ten APIs can still be perfectly manageable with direct Fabric REST ingestion if Fabric is the main consumer and the APIs have straightforward authentication, pagination and rate limits.
Microsoft’s Fabric REST connector already supports pipeline Copy activity with pagination and authentication options including service principal.
I would bring APIM in when you need an API management layer rather than just an ingestion mechanism - for example centralized authentication, throttling, quotas, policy enforcement, transformations, monitoring, or multiple consumers sharing the same APIs.
For cost/performance benchmarking, I would compare:
- request volume and concurrency;
- payload size and pagination;
- API rate limits;
- retry/error rates;
- end-to-end ingestion duration;
- APIM tier/capacity required;
- number of consumers using the APIs.
Microsoft’s API Management cost guidance is useful here because APIM cost depends on the tier and capacity/request model rather than simply the number of APIs.I would therefore start direct unless there is a clear governance or shared-platform requirement, then introduce APIM when the API layer itself needs centralized management.
AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.