Summary
Why I Turned My Synology DS218J Into a Dedicated Storage Device—and Why My Homelab Is Better Because of It
My Synology DS218J wasn’t a purchase I carefully researched or planned. I actually won it through a random YouTube giveaway, and for quite some time it remained unused in storage while I was studying Computer Science and Engineering and living in a dormitory.
After graduating and returning home, I finally powered it on. Like many new NAS owners, I initially used it as both a storage device and an application server because nearly every guide recommended that approach.
That changed when I built my homelab server using an old Dell laptop. From that point on, I completely abandoned the idea of using my NAS as a server. Instead, I assigned it a single responsibility: storing data. Every computing task moved to the server.
Separating storage from compute ultimately transformed my homelab into a safer environment for experimentation while also giving me greater peace of mind whenever something went wrong.
My NAS Runs Nothing—And That’s Completely Intentional
I Ignored the Advice That Told Me to Do More With My NAS
Over the years, I read countless NAS buying guides. Nearly all of them recommended installing Docker, media servers, photo management platforms, download clients, and a variety of additional applications.
For several years, I followed that advice without questioning it. I installed almost everything I came across, including Synology applications such as Photos, Drive, Office, and Download Station, along with third-party software like Plex and ElephantDrive.
When I eventually discovered Docker containers, I wanted to adopt them as well. However, the DS218J was not one of Synology’s Docker-compatible models. At one point, I even considered replacing it with a more powerful model such as the DS225+.
Several years later, my NAS runs zero Docker containers and almost no Synology packages beyond the essential services required for storage and file sharing. It has become nothing more than a dedicated storage appliance.
Ironically, after building my homelab server from an eight-year-old Dell laptop, upgrading my six-year-old NAS no longer felt necessary.
An Old Laptop Became My Compute Server
The laptop is a Dell Latitude 7480 equipped with an Intel Core i5-6300U, 8GB of RAM, and a 256GB SSD. I later upgraded it by adding another 4GB of RAM from a different retired laptop.
Today, that machine runs more than 15 application stacks with over 20 Docker containers powering my home infrastructure.
Its workloads include:
- Media and content services such as Jellyfin, Immich, and Nextcloud
- Networking services including AdGuard Home, Unbound, Tailscale, and Newt (Pangolin connector)
- Monitoring tools such as Uptime Kuma, Beszel, and Speedtest Tracker
- Utility applications including qBittorrent, Scrutiny, Vaultwarden, and Omada Controller
For persistent storage, every one of these services mounts directories from the NAS through SMB and NFS.
This decision wasn’t driven by hardware limitations. My NAS was still capable of running applications like Photos, Drive, and Plex without issues. I could have upgraded to a Docker-enabled model such as the DS225+, but that wasn’t the goal.
I simply wanted my NAS to perform one job—and only one job: storage.
That single design choice made my homelab significantly simpler than any hardware upgrade could have achieved.
I Constantly Break My Server—But My Storage Never Notices
Separating Storage From Compute Is More Valuable Than It Seems
Because of the nature of my work—which I genuinely enjoy—I regularly experiment with new technologies.
One day I might evaluate multiple DNS servers. Another day I may deploy a self-hosted replacement for Tailscale’s cloud control plane. I routinely test MCP servers, self-hosted alternatives to Cloudflare Tunnel, and entirely new Docker stacks.
As a result, my Debian server evolves almost every week. It functions not only as my homelab but also as the primary testing environment for new ideas and services.
Despite using Linux for many years, I still approach Debian upgrades cautiously.
The difference is that my data no longer resides on the compute server.
Because my storage exists elsewhere, I don’t worry about experimenting aggressively. Rebuilding containers, deleting entire stacks, or rolling back failed deployments has become routine rather than stressful.
Whenever something fails—or an experiment simply doesn’t deliver the expected results—I can restore the working environment by reconnecting to the same NAS-mounted storage.
Keeping data separate significantly reduces the mental burden associated with experimentation.
I now view my homelab as two completely independent systems:
- One system stores data.
- The other runs software.
Real-World Examples of Why This Approach Works
Last month, I performed extensive testing across several DNS solutions.
For years, I relied on Pi-hole together with dnscrypt-proxy.
Later, I migrated to AdGuard Home because it allowed me to replace a two-container configuration with a single container while offering native DNS-over-HTTPS (DoH) support.
After that, I switched to Technitium DNS Server, which felt more like a full-featured DNS server than simply an ad blocker with DNS capabilities.
Today, I’m once again testing AdGuard Home, this time paired with Unbound for native recursive DNS resolution.
These experiments are conducted across my entire home network rather than on a single device, meaning every connected device depends on whichever DNS solution I’m evaluating.
The process remains stress-free because no matter what happens to the compute server, the data stored on my NAS remains unaffected.
When Automatic Updates Went Wrong
A more practical example involved Watchtower.
I originally deployed it while writing an article, ended up liking it, and decided to keep it running.
Last week, both my Immich and Nextcloud deployments suddenly stopped working.
During troubleshooting, I discovered that Watchtower had automatically upgraded Immich to version 3.0.1 and updated Nextcloud to a schema-breaking release without my knowledge.
I successfully restored Nextcloud using:
docker exec nextcloud php occ upgrade
Immich, however, still required a PostgreSQL image migration.
Even so, the entire incident remained isolated to the compute environment.
All of the data stored on my NAS—including the data used by both Immich and Nextcloud—remained completely untouched.
This is only one example among many.
Every time something breaks, the most important fact isn’t that I eventually recover the services. It’s that my NAS remains entirely unaffected.
It has no awareness of the chaos happening on the compute side.
Because storage and compute are fully separated, almost every issue remains isolated to a single layer of the infrastructure.
My NAS Became Boring—Which Was Exactly the Goal
Everything else inside my homelab changes almost every week.
My NAS, on the other hand, has required virtually no attention for months.
To be honest, I rarely even log into DSM anymore.
Occasionally, I’ll open Jellyfin and play a movie. If playback works, I know the NAS is functioning exactly as expected.
In fact, I can’t even remember the last time I signed into the DSM interface.
The only reason I’ll open it today is to capture screenshots for this article.
When I describe my NAS as “boring,” I mean it as a compliment.
Because storage is isolated from compute, every software experiment carries minimal risk.
I never check the NAS CPU usage.
I never wonder whether deploying another container will overload it.
It quietly serves files to the server—and that’s precisely what I expect a storage appliance to do.
A Dedicated NAS Still Makes Sense
None of this suggests that running applications directly on a NAS is a bad idea.
Modern Synology systems—and many competing NAS platforms—run Docker containers without difficulty.
My DS218J couldn’t do that even if I wanted it to, but that’s not really the point.
My workflow revolves around frequent software changes, new Docker stacks, and intentionally breaking things to understand how they behave.
For an environment like mine, the most reliable component is the one that performs a single, predictable task.
The more I experiment with software, the more I appreciate hardware designed around one responsibility.
Sometimes the Boring Solution Is the Best One
At some point over the past year, after moving all compute workloads to my dedicated server, I stopped thinking about my DS218J as an underpowered NAS that needed replacing.
Instead, I began seeing it as a simple storage device with one reliable and predictable responsibility.
In my homelab, that simplicity isn’t a limitation—it’s an advantage.
Not every homelab needs an all-in-one solution.
Sometimes, the least exciting architecture is the one that delivers the greatest long-term reliability.