From Martin Casado on twitter:
Take standard compute in the cloud ...— martin_casado (@martin_casado) August 15, 2019
- Code written in high-level language
- Language libraries
- Language runtime
- Userspace OS libraries
- Guest kernel
So many redundant memory managers and networking stacks alone ...
The entire thread is fascinating, and his point is quite valid: How does the way we architect, code, and deploy applications in the cloud affect the cost in terms of computing power, as well as in margin?
For starters, the redundant stacks are heavyweight, and to some extent, both containers and serverless are meant to address this. But even those are not without cost. It’s not hard to imagine the real cost in cloud computing comes from the multitenancy.