Skip to main content

Do you actually need one?

Before spending anything, it's worth being honest about whether hardware solves your problem. Sometimes it does. Often the real constraint is somewhere else and a purchase just moves the frustration.

The signals that your current machine has genuinely stopped being enough​

These are the ones I'd act on:

  • You're out of memory, repeatedly. Not once during a heavy module, but routinely. You're shutting down one machine to start another and it interrupts what you're actually trying to learn.
  • You want things to stay running. A lab you rebuild every time you close the lid is fine for lessons and useless for anything that needs to observe over time: monitoring, log collection, backup schedules, certificate renewals. Half the interesting failures only appear after a few days of uptime.
  • You want to break things properly. There's a category of learning that involves genuinely wrecking a machine, pulling a disk, or cutting power mid-write. Doing that on the laptop you also work on is a bad idea.
  • You want to practise the physical layer. VLANs on a real switch, cabling, a firewall appliance with actual interfaces. You can simulate a lot of this, and at some point simulation stops teaching you the thing that goes wrong in a datacentre.

The cheaper answers to try first​

Every one of these has solved the problem for somebody who was about to spend money:

  • More RAM in the machine you own. Frequently the single best price-to-relief ratio available, and often under $100. Check the machine's maximum first, and note that a desktop is usually both cheaper and easier to upgrade than a laptop.
  • An external SSD. If your complaint is disk space rather than memory, this is a fraction of the cost of a server.
  • Shut things down. Genuinely. The course's Tier 1 works because machines that aren't in use are powered off. If everything you own is running all the time, you have a habit problem rather than a hardware problem.
  • Cloud, briefly. For a specific experiment that needs more machine than you own, a few hours of a large cloud instance costs less than lunch. It's a bad permanent home for a lab and an excellent temporary one.

Two ongoing costs, and why the design is small​

Electricity is the one that continues forever. CyberRack is designed to sit around 125 watts idle, which would be roughly 90 kilowatt-hours a month, about $13.50 at $0.15 per unit. Work out your own rate, because that number is the difference between a lab you keep and one you quietly switch off. Measure the real draw with a plug meter once it's running, since a design target is not a bill.

Time is the other. Hardware you own is hardware you maintain. That's part of the learning, and it's also evenings not spent on the curriculum.

Everything else people warn you about (noise, heat, whether it's welcome in the room) is what the small form factor already buys you. Used enterprise rack servers are tempting because they're cheap, and they're cheap because they sound like a hairdryer, draw several hundred watts, and were designed for a room nobody sits in. CyberRack is built from mini PCs specifically so the lab can live where you do. If you go a different way, that's the trade you're making.

The honest recommendation​

Finish the course first, or at least get well into it.

By Module 12 you'll know which parts of infrastructure you actually enjoy, and that changes what you should buy. Somebody who discovered they love detection engineering needs memory and storage for logs. Somebody drawn to networking needs a managed switch and a firewall with real interfaces. Somebody heading for platform engineering wants nodes for a cluster.

Buying before you know which of those you are is how people end up with expensive hardware that doesn't fit the thing they turned out to care about.

If you've read all that and still want one, the next page is the order I'd buy in.