Hello @ilkka.kosunen@gmail.com ,
Azure Time Series Insights can be seen as a data lake with data stored in Apache Parquet format and with a visual explorer on top of it.
It is capable of storing years of data and makes it queriable to you using the explorer (or via API) in seconds.
Yes, you can use Azure Time Series Insights with almost any data source (via an EventHub as event source) which matches a qualified Time-series-ID and optionally an event-source-specific timestamp property name.
There is even a usecase describing ingesting data from eg Stream Analytics of Azure Functions (thus the warm path output).
Time Series Insights (gen2) has its own storage, based on a regular Azure Storage data lake (gen 2). So you do not have to store data yourself, this is done in that data lake. Technically, you even have access to the raw Apache Parquet so you could do additional querying outside TSI. Keep in mind that data is mended to be readonly, you can not alter it or delete it.
You can start querying using the Visual Explorer and save and share certain views on the data. Of course, this is done by hand.
If you want to query via some programming, most of the time, the TSI Rest API is powerful already for custom queries.
There are some small costs involved, every time you access this 'cold store'. If you query really a lot of time, this can add up. TSI also offers a 'warm store' (see it as data kept in memory, the data will always arrive in the cold store). You pay upfront for this but querying is free (up to 31 days).
Personally, I like it a lot because TSI gives me direct visual insights into my data lake thus the long-term stored raw data.
An introduction to TSI can be found here.
--If a reply is helpful, please Upvote and Accept it as the answer. This will help others with the same question--