Hacker News new | past | comments | ask | show | jobs | submit login

> Most of our conversation focuses on scaling the database.

I think one of the emerging principles behind modern micro-service design is to break out your data model into separate services that hide the database. You can then publish data changes to an event stream. This can help avoid requests coming back to a database. I think this is a better approach compared to heavy caching (e.g. redis / memcache).

I definitely agree that micro-services aren't a silver bullet that solve every engineering problem that exists but the hyperbole of 150 micro-services is a straw man argument.

My main annoyances with Kubernetes/docker systems I have encountered is stability of the cluster and visibility into the health of pods. Both of these issues were the result of my org deciding to build our own Kubernetes from scratch and this has turned out to be a significant task.

If I was starting my own company and wanted to develop using micro-services I would probably use one of the existing off-the-shelf container cloud service providers (e.g. Amazon/Google/Microsoft). I think that is a better approach than "build a monolith then lift-and-shift into micro-services later".




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: