Imagine the potential of Infrastructure from Code (IfC) alongside IaC - it could change how we handle infra deployments in the future.
In addition to efficiency in deployment, IfC can also help ensure that all deployments are compliant with regulatory requirements, security policies, and internal best practices.
Honestly? I understand the value for certain vendor, but the run OUT of the cloud, or someone else computer, built like a mainframe, meaning something already surpassed decades ago, is started.
We need to get back in control of the bare metal and the basic infra on top of it, not another wrapper on some wrapper on some other wrapper. Deployments of the future are on personal iron, via some LOM with declarative systems like NixOS or Guix System, not on some cloud vendor APIs.
I found myself wanting something like IfC while using terraform for the first time. I wrote 40 sloc to deploy a web app in under an hour, then shut it down cleanly. Everything I did could be tracked in git -- great!
Now with a couple of tweaks I could deploy to another cloud, right? Well, actually, Terraform is a thin (stateful) wrapper around vendor APIs which may not have functional equivalents elsewhere. Even for a simple hello world app, I had to completely rewrite my states to account for how compute nodes are provisioned.
In other words, even when using terraform, I was forced to use two entirely different languages to describe what I actually wanted, which was a machine with X CPU and Y RAM and no more than Z ms of latency to my target audience for less than $A/mo.
6 comments
[ 3.4 ms ] story [ 21.2 ms ] threadIn addition to efficiency in deployment, IfC can also help ensure that all deployments are compliant with regulatory requirements, security policies, and internal best practices.
We need to get back in control of the bare metal and the basic infra on top of it, not another wrapper on some wrapper on some other wrapper. Deployments of the future are on personal iron, via some LOM with declarative systems like NixOS or Guix System, not on some cloud vendor APIs.
Now with a couple of tweaks I could deploy to another cloud, right? Well, actually, Terraform is a thin (stateful) wrapper around vendor APIs which may not have functional equivalents elsewhere. Even for a simple hello world app, I had to completely rewrite my states to account for how compute nodes are provisioned.
In other words, even when using terraform, I was forced to use two entirely different languages to describe what I actually wanted, which was a machine with X CPU and Y RAM and no more than Z ms of latency to my target audience for less than $A/mo.