Understanding the 4-Minute SLA in Azure Data Factory Activities

SLA of the Azure Data Factory service

This might be the first time you’re hearing this, but the Azure Data Factory service does NOT GUARANTEE an immediate or even a few seconds’ delay in starting any single activity in your pipelines! In fact, the guarantee for execution time is just 4 minutes, and it’s mentioned both in the FAQ section of the service and in the official SLA summary document for all Azure services (November 2023).

So, if your solutions MUST guarantee minimized execution time, then from the perspective of the service’s SLA, you also need to consider this in your risk assessment. Because in the worst-case scenario, any activity (if activity, pipeline execution, running copy activity, foreach loop etc.) might start with a delay of up to four minutes, and this is completely acceptable.

Does this mean that the time is always so long, and now I need to look for alternatives?

Of course not! Data Factory is a fantastic service that supports solutions for a huge number of clients. Every day, countless activities are orchestrated, and oceans of data flow in. It facilitates importing data into Azure in ways we could only dream of in SSIS. But the somewhat bitter price for this scalable, distributed, and multitenancy solution is the nondeterministic execution time of our activities. I will cover this topic in more detail in my upcoming articles 🙂

Isn’t this 4min’s SLA for all activities combined?

Here, it remains to dispel any doubts about the literal understanding of the scenario described in the SLA. To be sure, I asked Uncle GPT for his interpretation of these provisions. It seems that we understand the matter in the same way:

Thanks for reading!


Four minutes is also a very significant amount of time, which is the subject of a certain song. 😉

Leave a Reply