Self-Hosting

You can build a working 10-gigabit home network for around $200 by combining a budget 5-port 10GbE switch with Cat6a cabling and affordable PCIe NICs. The result saturates NVMe-backed NAS transfers at roughly 1,100 MB/s - about ten times faster than gigabit Ethernet - without enterprise-grade hardware or specialized wiring. Most of the savings come from buying NICs secondhand and sticking with unmanaged switches.

When Gigabit Becomes the Bottleneck

Gigabit Ethernet tops out at around 112 MB/s of real throughput. A basic NVMe SSD reads at 3,500 MB/s or more. That means your network is roughly 30 times slower than your storage - and for anyone running a NAS, this gap shows up fast.

That retired ThinkPad or Dell Latitude in your drawer beats a Raspberry Pi 5 on raw CPU, draws about the same idle power, and ships with a free UPS: its own battery. Install Ubuntu Server or Debian , enable SSH, and close the lid. You now have a silent homelab node that runs Docker, file shares, or home automation at zero hardware cost.

This guide covers every step: picking the right laptop, prepping the hardware, and installing a headless Linux distro. It also walks through hardening it for always-on use , running services with Docker Compose, and keeping the box stable for years.

A Unified Kernel Image is a single PE-format EFI binary that bundles the kernel, initrd, kernel command line, OS release metadata, and an optional splash into one file that UEFI can execute directly and Secure Boot can sign as a whole. On a 2026 Arch, Fedora, or Debian system the canonical toolchain is ukify from systemd 256+ to assemble the image, sbctl to enroll Secure Boot keys and sign it, and systemd-boot to auto-discover the signed .efi from /boot/EFI/Linux/. The result is one signed file per kernel that replaces a mutable loader/entries/*.conf, a GRUB installation, and separately signed vmlinuz and initrd files.

To batch migrate GitHub to Gitea, migtea gets you past one small limit: Gitea’s built-in migrator moves exactly one repository per submission. migtea is an open-source terminal tool I built that moves your whole account in one pass, wikis, LFS, and issues included. Here is how it works, screen by screen.

Everything here was tested against Gitea 1.26.4, gh 2.96.0, tea 0.12.0, and uv 0.11.2.

Why won’t Gitea migrate all your GitHub repos at once?

Gitea has a built-in migration page, but it moves only one repository at a time. You paste one GitHub URL and a token, click Migrate, and wait. The repo arrives complete, with its wiki, LFS objects, and issues. Then you start over with the next one. For a hundred repos, that is a hundred rounds of the same form. migtea does them all in one go.

The WD Black SN8100 (Gen5, TLC, up to 14,900 MB/s) is the best overall NVMe SSD for homelabs in 2026, while the WD Black SN7100 (Gen4, TLC) offers the best value for most builds. Avoid QLC drives for write-heavy NAS workloads - TLC endurance is worth the small price premium when your data matters.

Picking an SSD for a homelab is different from picking one for a gaming PC. Your homelab drives run 24/7, handle sustained write workloads from VMs and containers, and need to last years without surprise failures. The wrong choice - a cheap QLC drive in a write-heavy Proxmox setup, for instance - can burn through its endurance rating in under a year. This guide covers which drives are worth buying right now, what to skip, and how to keep them healthy once they’re installed.

The best eGPU enclosures for Linux in 2026 are the Razer Core X V2 ($349, Thunderbolt 5, 80 Gbps) for maximum bandwidth and the Sonnet Breakaway Box 750 eX ($349, Thunderbolt 4) for proven Linux reliability. Thunderbolt 5 enclosures have finally closed the bandwidth gap that made external GPUs feel like a compromise, and Linux kernel 6.12+ delivers stable hot-plug support that actually works.

External GPUs spent years as a niche curiosity - the bandwidth penalty was too steep, driver support too fragile, and the cost math rarely made sense. That calculus has shifted. If you run GPU workloads on Linux - local LLM inference, Stable Diffusion, CUDA development, PyTorch training - an eGPU setup now gets you 85-95% of internal PCIe performance depending on the workload. This guide ranks the enclosures that work best on Linux, walks through the setup process, and sets realistic expectations with actual benchmark numbers.