re: You Can Do it in SQL, Stop Writing Extra Code for it VIEW POST

FULL DISCUSSION
 

This is all good, but it leaves out a major point: Pulling directly from a database (or a filesystem) and exposing that to the consumer is not really scalable.

Sure, it works fine in a single application with its own database, but imagine 50 different webpages and services working off that same data. If one service makes a change to the database, then everyone needs to make changes to their code.

But, if everyone depends on a data accessor library (or service, or whatever), only the accessor needs to change! Everyone else's contracts will remain the same.

True, you're probably not going to swap out the entire underlying store all at once, (It happens, though... Especially early on in the lifecycle of the datastore.)

That being said, everything above is still applicable! Databases often are highly optimized for pretty much everything you pointed out (though sometimes full-string operations slow things down a LOT... I'm looking at you, Bob. I saw you put that LIKE statement on a multi-million row join. I warned you. I warned you about the stairs.)

I'm looking at a bunch of BigQuery stuff right now at my current job, and I keep seeing these queries that over-use with(). It's a simple join, Karen! It's not even a NOT IN or fancy grouping! But BQ is usually fast enough that this doesn't even matter...

 

Appreciate your views. As the example mentions it is a "microservice" so it is assumed that the DB is accessed by only one application. If you have 50 applications accessing one DB there is a bigger problem to solve IMHO. Not using SQL well is a different problem like you pointed to Bob for the like query. That could even run kinda OK if the indexes are placed logically. Thanks!

 

I would think that a database would be the one common thing when doing Mico services. Do you really scatter the data around like you do service endpoints. (my experience using and mocking micro services may not be representative.

Keeping data central causes all the problems @jbristow mentioned.
And a "data accessor library" isn't a solution, it's just a bottleneck, in terms of flexibility if not even performance for your whole system.

Microservices should own their data, and communicate through APIs they expose, ideally through commands and events.

code of conduct - report abuse