Project Goliath (1/3) - The Plan

Introduction: Outgrowing My First Network Attached Storage (NAS)

My self hosting and homelab journey originally started with my 1st NAS. Not a single switch, mini-pc, power distribution unit, or Raspberry PI. A old, decrepit HP Z440 that still has refused to give up. While it is a workhorse, it is not power efficient, quiet, or sexy.

The Z440 is equipped with 3 10TB HDDs, and I cannot add any more capacity to meet my ever growing storage needs. At 9TB of capacity, ~50%, has been already used and it was time to plan its successor. This wasn’t just about adding more space; it was about designing a server that could handle my entire footprint for the next five years.

Addtionally, when I think of that Z440 I think of spaghetti. It is a machine with Proxmox installed on the bare metal. That Proxmox instance serves as a plate for twists, turns, and knots of Virtual Machines, Containers, and neglect that I don’t know if I care to fix. “Why fix it if it ain’t broke?” I would be a much happier camper if my services were physcially separated from my NAS. I am just one failing power supply or motherboard away from losing access to most of my services. A part of me worries, and another part of me pretends to be ignorant. A plan for the future must be created and executed.

The Philosophy of the Build

Before a single part was ordered, I laid out the guiding principles:

ZFS is King:

The entire build revolves around the needs of ZFS (Zettabyte File System). ZFS is an advanced system for storing your computer files that’s focused on keeping your data safe above all else. To fully leverage ZFS’s data protection, the build needs an abundance of ECC RAM, a powerful multi-core CPU for checksums and compression, and direct access to lots of drives.

Imagine your computer’s hard drive is a library full of books:

• A normal file system is like a lazy librarian. It puts a book on a shelf and trusts it will be perfectly fine forever. Over time, a page might get torn or the ink might fade (“bit rot”), and the librarian would never know until you try to read it and find it’s gibberish.
• ZFS is like a super-meticulous librarian. When it stores a book (your file), it also writes down a unique description of it (a “checksum”). Periodically, it goes back, checks the book against the description, and if anything has changed—even a single letter—it knows the book is corrupted.

If the ZFS librarian finds a corrupted book and you’ve given it a backup copy (by using more than one hard drive), it will automatically and instantly replace the damaged one with the perfect copy. You won’t even know it happened. This prevents your photos, videos, and documents from becoming silently corrupted over time.

Instead of dealing with separate hard drives (C:, D:, E:), you just throw all your drives into one giant “storage pool.” ZFS manages them all together, giving you one big space to work with. It’s like combining several small water buckets into one giant swimming pool. This makes managing your storage much simpler and more flexible.

This system makes a lot of sense for my intended use case. I have a lot of media that I do not want to lose as it would be incredibly hard to get it back. For example, I have a copy of a popular movie about hobbits moving a ring to a volcano, but it is a special 35mm scan. A 35mm scan is a high-quality digital file created directly from the original physical film reel a movie was shot on. Think of it as a perfect digital photograph of the master copy, frame by frame. This process captures the movie’s authentic theatrical look, including its unique grain, color, and rich detail. Unlike a standard Blu-ray, which is often digitally processed for home release, a 35mm scan is a pure, archival version. This makes it a rare and often superior copy that is irreplaceable for film enthusiasts.

ECC for Extra Integrity

To this day, I am shocked people still consider a NAS without ECC ram. I am going to stack both ZFS and ECC ram, to try and completely elimate the possibility of bit rot.

ECC (Error-Correcting Code) RAM is a special type of computer memory that acts like a meticulous proofreader for your system. Its job is to detect and correct the most common kinds of data corruption that can happen inside the memory itself.

Think of your computer’s memory as a busy scribe’s temporary notepad:

• Standard RAM is a fast but fallible scribe. As he’s writing down data, a random error—caused by anything from cosmic rays to a slight power fluctuation—can make him write “The quick Crown fox” instead of “The quick Brown fox.” He won’t notice the mistake.
• ZFS, the super-librarian, then takes that notepad. It sees “The quick Crown fox,” and dutifully saves that corrupted sentence to the permanent library (your disks), complete with a perfect description of the corrupted version. ZFS did its job perfectly, but it was given bad data to begin with. This is how silent data corruption begins.
• ECC RAM is a self-correcting scribe. When that same random error happens, a tiny alarm bell goes off. The ECC RAM instantly detects the mistake, corrects “Crown” back to “Brown” on the spot, and then hands the perfect data to ZFS.

ZFS’s legendary data protection begins once the data is ready to be written to disk. It cannot protect you from errors that happen in RAM before that point. Using ECC RAM ensures that the data ZFS receives is clean, making it the bedrock of a truly stable and trustworthy storage system.

Data degradation, bit rot, or whatever you want to call it happens. More than you think. Hell, while procuring components for Goliath my gaming PC had corruption within the operating system files. This forced me to completely reinstall Windows.

Enterprise Heart, Homelab Budget:

The real magic of this build, and the secret to getting mind-blowing performance on a reasonable budget, is harvesting retired enterprise gear. We’re talking about components that were once the heart of a corporate data center, costing thousands of dollars just a few years ago, that can now be yours for less than the price of a new video game.

Many people hear used parts and immediately question their reliability. It’s a fair concern, but it stems from thinking about used consumer hardware. A five-year-old gaming graphics card has likely lived a hard life. A five-year-old server CPU? It’s just getting warmed up. In fact, many components that I will be using are designed to run 24/7/365. The materials are better, the validation is more intense, and the failure tolerance is much stricter.

Why does this powerful hardware even become available? Big tech companies and financial institutions operate on strict 3 to 5 year refresh cycles. They replace entire server racks not because the equipment is failing, but because it’s more economical for them to upgrade to the latest tech for power and density reasons. The perfectly functional, high end hardware then floods the second hand market. It’s the ultimate billion dollar hand me down.

A Foundation for the future:

Much like a house this isn’t a temporary thing. I am starting with a strong foundation and framework that can support future addtions and renovations without having to tear the whole thing down.

For example:

• The chassis that I have selected has 24 drive bays, empty rooms ready for future capacity needs.
• The motherboard should have extra empty DIMM and PCIe slots as expansion potential. If I ever need a fiber network interface, I’ll have a spare x8/x16 slot to plug it into.
• The power supply will have extra wattage “headroom” to handle the power draw of expansions and addditional spinning drives.

This is much better to the alternative of having a “dead end” build. Pre-built NAS devices or complete solutions for small businesses often force you into a corner. They don’t have extra space for additions, or any ability to replace components with a better one if the need arises. If you want to make a major change, a full replacement would be neccessary.

This build will also be intentionally overpowered for its initial role. With the initial BOM already in mind only thing I can forsee upgrading is adding that fiber network interface to achieve 10Gb or even 40Gb networking!

Summary

And there you have it. The complete blueprint for Project Goliath. This is more than just a plan to replace an old server; it’s a declaration of war against data loss, inefficiency, and dead end hardware. By embracing the obsessive data protection of ZFS and ECC RAM, and leveraging the immense value of the enterprise hardware market, the vision for a powerful, reliable, and scalable home for my data is finally clear.

The philosophy is defined. The requirements are written in stone. The only thing missing… is the hardware itself.

Join me in the next post, where I transition from planning to procurement. We’re going treasure hunting for the data center cast-offs that will form the heart and soul of this storage beast.