4 ms·
> microservices etc are not helping either. When done well they are supposed to help, but in reality I only see systems that can work when all microservices are
by robertlagrant 5y ago
> microservices etc are not helping either. When done well they are supposed to help, but in reality I only see systems that can work when all microservices are up and responsive; if one is down, the entire thing is dead. Why didn't you make a monolith? Now you have brittle all over the place and devops with 247 stress.
You could make a monolith where the UI keeps working even if some of the monolith's endpoints don't work, and you can make a microservices architecture where the UI still doesn't cope if a microservice is down.
I think the pros of microservices are:
- deploy smaller - you only update the parts of the system you need to when you modify something.
- different technologies - you can use Ruby here and Go there if you like; very un-locked-in and you can maximise the value of any libraries you have. E.g. if you have a number crunching bit of your app you could make a Python microservice with numpy etc installed.
- independent data stores - pro and con, of course, but it's nice if you can decouple bits of your system and again use mongo here, postgres there if you need to
- as an microservices-based application grows in scope, the number of engineers working on different bits of it can scale, as they can deliver independently. It's harder to scale engineers working on the same codebase