but it does seem that most homelabs are big server racks with at least two CPUs and like 20 cores in total.

That’s just the people that post a lot online about it.

And these racks usually start with “I found this super cheap rack server on eBay” and not an actual need.

I have built a tiny setup in a 10" 6U rack.

• Os6250-24p poe switch
• 4x soquartz blade (4-8GB ram, 500GB nvme each)
• 2x external HDD raid enclosure

It runs a k3s cluster (on dietpi) with a couple of services:

• searxng
• nextcloud
• git
• borgbackup
• my website

I plan to add more stuff as the need arises. And the rack has 4 more empty slots for further blades.

The total cost so far:

• ~100$ per blade + soquartz (~400)
• ~45$ per nvme (~200)
• ~75$ per hdd HDDs (~300)
• ~50$ for the switch
• ~100$ for the case and random stuff

Total ~1050$

Mostly it is a fun learning experience, since I wanted to get better at devops. Also now I can easily use it to deploy my hobby projects (like the website) in a docker container, which makes it more fun than subscribing to a recurrent payment for a server host. Also, since I have it in my home, it doesn’t matter if there is an internet outage, I can still access the backups and the data on my nextcloud. As a result my computers are very lean and I can easily replace them without having to think about moving important data from the old machine.

I wouldn’t want to get rid of it now.

Whether you need a high power homelab on the other hand is up to your budget and intent. I see it similarely to cars or bikes.

But I think a prebuilt NAS is way more expensive, less flexible and really not worth it unless you really dislike the tinkering aspect.

The use case is that I want beefy server hardware just because.

My take (having neither but building a NAS in the background of other jobs) is that if you don’t need the rack, don’t buy the rack. If you already have a NAS and you really want to play with the power that a rack would give you, go for the rack. If you don’t need it don’t buy it, simple as.

I started with a basic as shit NAS and was happy with it.

Then i wanted Hardware Acceleration for my Plex Server because i wanted to stream high resolution content when i was out of the house.

I then rebuilt my old Gaming Rig into a server.

After i realized that i now have much more power to use i started to host a bunch of services; AdGuard, Plex, Sonarr, Radarr, Prowlarr, Overseerr, Homarr, Lidarr, SabNZBs, Kavita, Kaizoku, HomeBox, HomeAssistant, Nextcloud, FoundryVTT, PaperlessNGX, Audiobookshelf, Romm and Whisper for my HomeAssistant.

That’s stuff i would’ve never even had the chance to host on something simple like a little NAS.

Oh and most homelabs are NOT racks with 2 cores… in my case, old gaming PC with Ryzen 5 2600X, 16Gigs of Ram and GTX1660 Super

deleted by creator

Having multiple sufficiently-powered virtual machines makes OS development really low friction. Though I’d personally go for a blade subrack instead.

So, convince me. Why should I get a homelab instead of a regular NAS?

Eventually it just might get out of hand and you end up with both. TLDR at bottom.

Serving out of the home and I’ve been racking uptime
Gotta gotta add more because I want it all
It started out with a switch how did it end up like this?
It was only a switch, it was only a switch

I started out in college with an old digital sign controller (2 core 4g RAM) and added a hard drive to a slot based on its OEM design and reflashed the bios back to OEM, what it’s value-added manufacturer never intended, to get the HDD to work. It was all the right price of free. That became my first NAS. Then I got the switch, then reflashed my router to OpenWRT for features, discovered it couldn’t route a gig to NAT anymore, so I did what any logical nerd would do: tried pfsense but ended up building my own Linux based firewall and everything server, also from the free electronics recycling bin. Then my old laptop got converted into a server

Once I got a job and a real budget, I started running into hardware limits. You can only run so much on hardware that is old enough to attend middle school. Out was the homebuilt router-everything server and digital sign NAS, in was the Synology DS1517+ and super micro 8 core mini server and in was the small rack. Settled on ESXi free for a hypervisor. Later, additional nodes were purchased and I upgraded from free to VMUG and added vCenter and vSAN (aggregate size: 1.5TB cache w/ 3TB capacity tier, which worked out to ~2TB total after RAID) to help manage it all. Windows Active Directory was built (had a leftover perma key from college) to coordinate my ever increasing VM count with the same password everywhere.

Timeskip forwards, and I have a pair of R720XD I got for cheap because I knew how to BIOS hack them off of their original Google Search Appliance firmware back to Dell stock, total vSAN capacity at ~8TB after RAID, the Synology is still alive but now rocks ~48TB after RAID, and one “loose” R720 with ~2TB storage and a then-new now-aging Intel NUC with 32GB RAM. All three R720 have 128GB RAM each, I have more switch ports than I’d like to count, and 15 minutes of battery backup added last year. The NAS backups up all devices with a minimum RPO of 1 week and a maximum RPO of 24h for critical stuff. RTO is 2-4h from event, and by golly that has saved my bacon a minimum of 6 times in 3 years. At this point I have more infrastructure redundancy and capacity than some medium businesses.

Could I live without it and it’s monstrous power bill? Yeah absolutely I could downsize that and save some cash. But where’s the fun in that? Every component I added to my system was done so with a very specific common goal: every piece of this monster was built to help me learn about something IT related. With this experience under my belt, I was able to confidently jump in to stuff at work with the mantra of “yeah I already know how to do this”.

Plus, as a side benefit, its a fun learning hobby. I have an absolute blast learning about all this technology and figuring out how to get the most bang for my buck when it comes to selecting software (paid vs FLOSS) and procuring hardware. All things you need to do in the real IT world. Sure I don’t get to play with the fancy spaceship servers that have multiple terabytes of RAM each like I do at work, but I don’t (yet) need a multi terabyte RAM chassis at home.

Summary and TLDR: Build something that solves a problem in your life. Photo video storage? NAS! Kids want better internet? OPNsense/OpenWRT firewall with a switch! – as two examples. But my chief most import prime directive rule: build something that makes you and your family happy.

with 1PB of storage

As data hoarder this is a goal to achieve :)

What’s the main advantage of having a huge server, compared to an average Synology NAS for example?

You can do much more things much more efficient. If you’re out for a NAS, get a DiskStation, even the simple ones with 2 slots are absolutely fine. Get a Seagate Exos X20 with 20 terabyte of storage, or 2 of them for a RAID, and you’re good to go. Or build something yourself with an external case and a Raspberry Pi.

Why should I get a homelab instead of a regular NAS?

Right now I’m using Docker and I’m hosting an ActivityPub server and a front-end (client) for it, a web server, a Minetest gaming server, a web Git platform and a custom application for some specific logging i’m doing. The same server also runs a reverse proxy and a Docker management platform. All on one older SOC machine with a recycled 1TB SSD. I couldn’t do this with just a NAS, even if the DiskStations are (were?) very “hackable”.

If you want to do anything more than just providing files via network, build a custom setup that fits your needs.

Definitely go for the beefy server. You can then run benchmarks just to flex

Something I’m not seeing here is many people in the cyber and info security spaces will have a homelab capable of complex configurations in order to mimic enterprise environments for their research.

This could be as simple as a single VM or as complex as multiple segregated networks to try to traverse.

Definitely go for the beefy server. You can then run benchmarks just to flex

Although 256gb of ram may not be enought to run matrix synapse.

Like others have said, it’s all about your goals. If you just want to selfhost some apps or have additional storage/local backups, by all means, keep it simple and efficient.

In my case, my homelab is justified because of my job. I’m a senior systems engineer who specializes in designing, deploying, and managing distributed infrastructure. So having a dedicated server room at home with a rack, 20A circuits, and HVAC allows me to more easily emulate certain environments, test out hardware on loan from vendors, experiment with new ideas or software solutions, and stay immersed in my craft.

Aside from the work-related benefits, in return for my higher than average power bill my home network gets to rival most corporations, and I can self host anything I (or my wife) wants.

While this setup is great for me, I would never recommend it for someone who would be better off with a NAS.

it does seem that most homelabs are big server racks with at least two CPUs and like 20 cores in total.

I’m not sure if your premise is correct, a lot of hardware discussion I see is people repurposing old desktops/laptops or running with sff/mff pc’s.

Why should I get a homelab instead of a regular NAS?

In case you want more than just storage like vpn services, hosting containerized applications, etc.

The big server is all about density and convenience. The thing will run many, many VMs without having to skimp on resources, and it will be easy to admin the VMs remotely.

I have plans to pick up a big workstation to replace the little desktops I have around, and it will be more convenient since getting as console on a VM is much easier. Servers might also have a BMC, which would help admin the server.