Hi all, I'm new to database design.

I'm collecting time-series high-resolution test data sets each computed/summarized into a simpler transactional set including its metadata. These are stored on a local server. I have experienced very few needs to join tables; however, I'm constantly using the time-series data to create other metrics to build ML models.

Each time-series dataset in CSV can be as big as 2.6MB while the transaction data is in kB size. I have over 20,000 sets of data.

What would be a database option/architecture with optimal performance? Relational or non-relational? What are the vendors for each option?

How do I determine if I should migrate to a cloud server? What are the recommended vendors?

What would be a database option/architecture with optimal performance? Relational or non-relational? What are the vendors for each option?

Any help or guidance or link to tutorial 101 would be highly appreciated!