
ML Ops in Production
Data Skeptic
00:00
Using Containerized Containers to Scale a Cloud Solution?
Most people still do homegrown. I think I am not actually sure why. Maybe because it's easier and our suspicion is because the people in charge are software engineers. And that's kind of how easy it is for them to at the end, we're machines. We like patterns. We basically match everything to that pattern. That's what they use in order to solve this problem. But actually it's not very efficient. Unless your model is very highly used, then it's actually more efficient to deploy multiple models to the same container. Otherwise, every model will get its own container. Every container will end up with you have a limit per node,. You end up with just spending a lot
Transcript
Play full episode