zet
Minimal home lab requirements for simulating enterprise HPC ML Ops
Every technologist needs a home lab, period. Not having one is the difference between getting hired and losing a job opportunity to someone else. I have seen this over and over again in people who lose their jobs and cannot get rehired. It is so significant I have had CTOs tell me directly, “I would hire someone with a home lab over multiple certifications any day.”
So how do you build the right one for your target tech career?
HPC ML Ops / SRE / SysAdmin
I think for people who haven’t learned all the infrastructure stuff even once they might want to focus on building a hardware lab of collected old hardware. Dell Optiplex is like 150 bucks and very powerful. Building a hardware lab requires power and space, however, and a lot of people in school or living in apartments aren’t gonna have that. I did this with a bunch of discarded computers in the garage after I had been married a while. I would keep my eyes peeled for people getting rid of “old” computers and breath new life into them as Linux servers. Keeping them up was part of the challenge.
I’ve had a few dozen hardware nodes in the past and spent a lot of time futzing with peculiarities that have nothing to do with the hardware we are on at work. It was good to learn PXE boot and networking and switch management and VLAN setup but for my current role those skills are okay to archive and sit on the shelf.
Backend Developer
The main difference between a home lab for someone focused on deploying APIs and backend infrastructure for an enterprise deployment model is hardware.
Ansible deployed Kubernetes clusters:
- prod
- dev
- inf