Hi! I have a question about the data modelling in Cognite Data Fusion.
We record data from thousands of sensors from a factory. Today we record them as time series in CDF and organized in the asset hierarchy by manually constructing the relationships.
I heared the data modelling toolkit is meant to replace asset hierarchies approach in future. Could you please elaborate a bit on how Cognite would suggest to approach our use case, if we started from scratch today.
Thanks!
Best answer by Glen Sykes
Hi Victor,
We are working on a feature right now to have Time Series as a first class data model type (meaning all the standard metadata in Time Series will be synced and present in your data modelling project) for up to 1 million time series instances at the time of launch. These limits will increase over time as we continue to make investments in storage scalability.
If you’re starting from scratch today, and are choosing between Assets API and Data Modelling, then providing your model can fit within the instance budget of Data Modelling (5 million nodes and edges per project), we would recommend you start there.
There’s nothing preventing the creation of an asset hierarchy like structure using data modelling, and you get the benefit of being able to use edges to enrich the hierarchy model where with Assets API, you have to use the separate Relationships API to do that and it comes with query limitations.
If the instance budget is insufficient for your requirements, let’s have a chat.
In particular we discovered that there is a limitation of how much metadata we can store for a time series / asset. Can this be overcome in any way?
Hi Victor,
We are working on a feature right now to have Time Series as a first class data model type (meaning all the standard metadata in Time Series will be synced and present in your data modelling project) for up to 1 million time series instances at the time of launch. These limits will increase over time as we continue to make investments in storage scalability.
If you’re starting from scratch today, and are choosing between Assets API and Data Modelling, then providing your model can fit within the instance budget of Data Modelling (5 million nodes and edges per project), we would recommend you start there.
There’s nothing preventing the creation of an asset hierarchy like structure using data modelling, and you get the benefit of being able to use edges to enrich the hierarchy model where with Assets API, you have to use the separate Relationships API to do that and it comes with query limitations.
If the instance budget is insufficient for your requirements, let’s have a chat.
Thanks Glen. If you have any link to the docs where I could read more on the parallel between Assets API and the Data Modelling, this would really help :)
Cheers,
Victor
We’re preparing material to talk about the parallels between our longer standing APIs such as Assets, Relationships etc, vs. Data Modelling, but the topic is quite large and complex as it touches many areas, such as how Time Series, Files and Events will play in a post-Data Modelling product architecture, and much of this thinking is emerging still.
We’re keen to get this right, so it might be a time before we can publish dedicated guidance, but a good start of course would be to follow the Data Modelling guides in our docs portal and where you see any shortcomings or problematic areas.
Happy in the meantime to provide specific guidance here as and when any questions come up.
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.
Check the
documentation
Ask the
Community
Take a look
at
Academy
Cognite
Status
Page
Contact
Cognite Support
