Skip to main content

To achieve real time data there are two possible ways.

 

1. Use the replication cockpit with a real-time extractor.

2. Use scheduled delta extractions as data jobs.

 

Option 1 obviously will be the recommended one. But what are downsides of option 2?

From my understanding the bottleneck of this whole process is loading the data model.

If that takes e.g., 10 minutes (which is a rather fast example), scheduled delta extractions, if scheduled more frequent or equal, should achieve the same thing like the replication cockpit.

So why would I use a real time extractor instead of that when I can achieve the same result with option 2?

Hello Luca! Enabling Replication Cockpit (and the resulting delta transformations) also enables delta loading in the Data Model. Essentially, moving to replication should enable the fastest possible response to source system changes. See this doc: Delta Data Model Loads (celonis.com)

 

However, if your requirement can be met without the need for replication, then certainly doing delta extractions will be easier to configure and maintain. Check your data growth expectations to determine if you will need replication in the future.

 

In my experience, "real time" analytics usually means "within a day/hour" of source system changes. In execution management, however, there are use cases that need a near real time response (ie: Action Flows).


Reply