Expected performance when linking to large tables


Hi all,

We are building a central data warehouse which stores/combines data from different source systems. Hence, here we have case, process, activity, etc. tables. However, these tables are large (GBs).

I would like to link Celonis to these tables but before I dare to do this I would like to know whether Celonis links to the source database or whether Celonis imports/copies the data into it’s own database?

Also: Where could I have found the answer myself? :slight_smile:


I believe detailed info should be inside data scientist course. At least it was when I took one (the real world, not online version).
In short it depends on the version/configuration of Celonis your are using. It will either run a huge select loading all even log data from your tables into Celonis application server memory and ignore your database after that, or share load between Hana database and appplication server.



you normally connect Celonis to your CELONIS Schema on your database server, where you have all of your transformed data like the Activity Table and the necessary Views.
When you load the data model, the current load gets cached on the application server where Celonis is installed.
So you would link Celonis to your tables.

Best regards,


Thanks for your reply Viana.

We did a PoC on a table containing 6 million cases and 38 million events, spread over 39 different processes. Performance (Celonis 4.4) is slower than usual but acceptable. Looking forward to expected increase in performance in version 4.5, as we expect more data since this is only a sample.