Comments URL: https://news.ycombinator.com/item?id=42267369

Points: 3

# Comments: 0

In a nutshell, any K8S deployment on-premises tends to be inherently optimized, saving a significant amount of time and resources. I have new servers and old servers in the same cluster, that is epic.

Modern FinOps often feels like a frustrating exercise: Should I choose 2x 2XXL instances or 4x 8XL instances? The conversation rarely focuses on optimizing software performance or database efficiency. Instead, the cloud has turned into a maze of cost centers, where it’s easy to get lost in ‘managing’ the cloud rather than building valuable products for end users.

* Less prone to human error. We have one well-secured, central firewall that only a few developers can access. So, even if a developer forgets to properly secure something downstream, it will still be protected by the firewall. One could argue that this is possible in the cloud, but managing VPCs, etc., introduces risks. There’s always the possibility of something critical being left outside the VPC. On-prem, there’s no way something can physically escape our ethernet cables.

* IAM and bucket management issues. Anything in the cloud is inherently exposed to the Internet and, in most cases, open by default. You need to manage countless IAM configurations.

* Physical inspection. We can actually look at our setup, and if necessary, visually inspect if a server is physically encrypted.

* Simplicity and transparency. Things are simpler and more straightforward: Storage is storage, a disk is a disk, and ethernet is ethernet. Canot stress how beatufill this is, even with 100 servers it easy to manage them than in the cloud.

* Modern open-source software. Modern open-source solutions have incorporated many smart features from the cloud, making on-premise setups more powerful and easier to manage.