Raspberry Pi 4: The Seed of a Homelab

Archive note: this is a backdated post, written years later while rebuilding this site. It’s dated to the moment it covers, but the hindsight is real.

The Raspberry Pi 4 came out in late June, and after weeks of reading reviews and telling myself I don’t need one, I want to write down why this particular board has me sketching diagrams of things my house does not currently require.

Why the 4 is different

Earlier Pis were brilliant educational toys that taught a generation Linux — and choked the moment you asked them to serve. The bottlenecks were structural: USB 2 everywhere, Ethernet hanging off the USB bus (so “100Mbps” shared lanes with your storage), 1GB RAM ceiling. Charming for blinking LEDs; miserable for anything with a database.

The 4 fixes essentially all of it:

Up to 4GB of real RAM — enough for an actual application stack, or several small ones.
True gigabit Ethernet on its own bus, plus USB 3 for storage that isn’t tragic.
A quad-core Cortex-A72 that benchmarks like a budget desktop from not-that-long-ago.
Same price class as always — the 4GB model lands around the cost of a decent dinner out for two. As infrastructure goes, it’s pocket change.

On paper — and per every review thread — this is no longer “credit-card computer for learning.” This is a tiny, silent, sips-power server. (Caveats duly noted: it runs hot enough to want a heatsink, and the early units have a USB-C power quirk with certain cables. Version-one tax.)

The dangerous train of thought

Here’s the sequence of thoughts the spec sheet triggers, which I suspect is the entire marketing strategy:

It could block ads for the whole network — there’s a project called Pi-hole that does network-wide DNS filtering, and every router in the house would benefit. It could run a file server — USB 3 plus gigabit means it would actually keep up. It could host my dev experiments — a always-on box for cron jobs, test deployments, that Telegram bot idea. It could run a media server for the family. And once one service works, the marginal thought is always: what else?

I recognize this slope from the way I fell into the terminal. It doesn’t end with one Pi. It ends with a shelf, a labeled power strip, and opinions about VLANs. I’ve seen the photos on r/homelab. I know what’s down there.

Why I think the slope is worth sliding

Beyond the toys, there’s a real professional argument forming in my head: running your own infrastructure is the cheapest systems education available. My day job lives on servers other people manage — increasingly abstracted away entirely. That abstraction is productive and it atrophies exactly the knowledge you need on the bad days: DNS, networking, storage, backups, what’s actually happening under the framework.

A ₱3,000-ish board where I’m the sysadmin, the SRE, and the user — where breaking things costs nothing but pride — is the lab fee for that education. Every service I self-host teaches a layer of the stack my CV claims I understand.

Filed intention

So, on the record, August 2019: I’m getting one. It starts with Pi-hole, because network-wide ad blocking is the demo that justifies the purchase to the household. Where it goes from there — file server, media server, the shelf, the power strip, the opinions about VLANs — I make no promises, which is itself a kind of promise.

Some purchases are gadgets. This one, I have a feeling, is a direction.