Question
How should personal lab, management, public service, and recovery paths be separated?
Trial log
NVIDIA DSX Air
Network simulation notes for learning before production infrastructure changes.
Purpose
Use simulation to make topology decisions safer.
DSX Air is treated as a learning and validation environment, not as a production VPS. The trial exists to understand how data center networking concepts behave before they become expensive or risky in live infrastructure.
Method
Run one topology at a time, record assumptions, change one variable, and capture rollback.
Output
Publish sanitized lessons; keep raw configs, screenshots, account details, and private topology offline.
Experiment queue
Start with primitives, not architecture theater.
01
Inventory
Record the lab topology, node roles, image types, resource limits, and stop condition.
02
Baseline reachability
Confirm management access, link state, IP addressing, and basic routing before adding overlays.
03
Segmentation
Model separate user, lab, public-service, and management paths with explicit trust boundaries.
04
Failure drill
Break one dependency deliberately, document blast radius, recovery steps, and time to restore.
First track
Cumulus Linux and EVPN foundations
The first network track stays close to primitives: switch roles, VLAN boundaries, routing intent, MLAG behavior, and EVPN/VXLAN concepts. The point is to understand what each layer is responsible for before composing them into a larger design.
- Document what problem each layer solves before configuring it.
- Separate underlay reachability from overlay policy in the notes.
- Record the failure mode for every convenience feature.
- Do not publish raw device configs until secrets, addresses, and environment details are reduced.
Current status
Trial state as of 2026-06-15
Public status
Sanitized trial plan published; no private topology or credentials exposed.
Next private action
Capture a minimal topology inventory and define the first baseline reachability test.
Stop condition
Stop the lab when resource usage, account state, or topology ownership becomes unclear.