I just listened to Justin Borgman on the opensourceunderdogs
podcast and found it really interesting to understand more about presto/trino, especially three things:
- To understand how they see themselves in the data mesh context
- To understand how they act as abstraction layer/decoupling component
- The general direction of future development.
I particularly like how Justin discusses the decoupling aspect in the sense that trino is kind of an analytical database without storage. No need to copy data around, stuff can happen at query time. Of course, the guys at firebolt
will have some things to say about whether that’s actually feasible or not, but I like the general direction.
I also like the project because it empowers teams to have an abstraction layer that essentially allows doing piece-by-piece migrations of data warehouses or data lakes; A concept I find very important. After all, I think you should be able to exchange all of the major components of your tech stack every 6–12 months.