Best practices for managing multiple datasources into 1 asset tree

[JDLom]

Hi,

We are running with a decentralized historian architecture having (Wonderware/ignition/Siemens/PHD) historians in every plant locally and sometimes even other SQL DB's, all containing their own framework, some pretty well structured, some not.

 

The goal is to provide the users a more structured way of finding all their relevant data of the business unit/ plants. The idea is to combine all these datasources into 1 tree

 

 

I'd like to learn what is the best way to manage this via datalab as these are the challenges that I see:

 

Some layers in the native frameworks are useless/not clear and I want to remove or rename them

I want to add a clear root on top of those trees (Business unit name / plant / rest of the tree)

Some plants also have other data sources (SQL) that I also want to embed into the right branch of the tree

Whenever the local OT engineers add a new signal to their historian, I'd like to have that signal in the new tree as well

 

Any tips or is this an idea to bury away?

 

 

[Patrick]

Hi JDLom - The current approach for a situation like this would be to leverage spy.search() to pull in the various structures and concatenate/manipulate the DataFrames to fit the structure you want.  Then follow the approach you took in one of your previous posts to push a tree based on DataFrame:

 

Regarding updating the tree when a new signal is added, you can perform a spy.search for the various sources to check for new signals, and then push new signals with the desired structure format.  This can also be automated by leveraging `spy.jobs`, which enables you to run the script on a schedule.

Hope this helps.