Hi everyone,
We are starting to implement Camunda enterprise-wide, via a centralized center of expertise and decentralize devs. It’s for a large company with hundreds of devs and thousands of services, we will have quite a lot of processes too, real time and long running.
We are using PCF as a PaaS solution for everyone developing micro services and exposing them and we also have Kubernets on two cloud provider.
As we want to build a centralized (but multi tenant) BPMS platform with Camunda, we are still hesitating between both approaches :
- Setting up Camunda directlin on PCF, as a normal java application and let the cloud bother with scaling and such, like the video and blog Bernd posted
- Going for the Kubernetes approach and build our own containers (runtime, external tasks optimize and elastic search) .
For the PaaS approach, one advantage would be the proximity to what our java developers are already doing for their normal API composition and such, especially in terms of CI/CD, methodology framework (springboot mostly) and such. The other advantage would be that it would be much faster to go to production than if we go the CaaS way.
However I have some worries about what DBs we can use and how flexible and efficient our scaling could be.
For the CaaS, we would have more control on everything, but much more work to do to (network, security, load balancing, helm, terraform, etc…).
I was wondering if others, here, had the same dilemna and what were the key elements that made them decide to follow one approach or the other.
Thanks in advance !