I am looking for a primary Nas system. Current need is at least 200tb system with double drive fault redundancy. I am not looking to host anything else on it. Needs to provide 12 users content. Storage will mainly be used for video backups. 10gb spidf backbone. Maybe 4tb of cache.
with 200TB usable, you’re lookin at 12x 20-22TB drives of data
we like 2 drives of redundancy per 6 so you’d be sitting at 16x 20-22 TB drives
a 9500 series Broadcom HBA
256 GB of ECC RAM (128 is feasible, but not best practice)
yeah, you definitely work with A/V alot
10 gig SFP NIC’s are cheap, Mellanox is the go to, don’t get an OEM branded one for production use
You can use PCIe to m.2 or preferrably U.2 as enterprise U.2 NVME drives have higher endurance and better thermal properties
1000+ watt titanium power supply to handle boot surge on 16 drives plus CPU, RAM, and NIC with ease.
3 Likes
This all makes sense. Any recommendation on mobo/cpu and case?
At least a 4U for hot swap ability and quieter fans
Are you running any VM’s?
So far your requirements are pretty tame compared to most of our deployments.
Make sure to get CMR drives
We deploy enterprise grade WD’s as they are being made in the HGST plant.
HGST was the industry leader for HDD reliability.
Though the only 22TB reliability information we currently have is just for HGST and no other brands with an AFR of 1.37%
Further reinforcing why we have 2 drive fault tolerance
Also I did mess up your calculation
You need 12 data drives
2 drive parity per 4 data drives
that’s 6 drives of parity for a total of 18 drives in production
That puts you over the 16 drive limitation of a 9500-16i card
Generally best practice to run 2 cards at this level so you can add drive ICE
Your internal constraints may dictate the decision here, but 16 drives should be minimum for a total of 4 drives of parity.
That’s where you want to be as the drives will likely all be from the same lot so when 1 fails, there will be more that will only be uncovered when rebuilding.
Given the pool size, an additional HBA and drives would GREATLY simplify rebuilding as few places have the means to perform data recovery on >12 drive arrays.
We recommend arranging the deployment so your parity drives are not all on one controller / port as that is a failure we see fairly regularly where just 1 port of the RAID card is dead.
1 Like
We use proxmox on all systems in our server rack. The current system is similar to what you have described, but is a mix of drive sizes on a 5700x cpu. No vms are run on our Nas system currently, we have 5 other systems for that. Thanks for your input on this.
1 Like
If you asked this question in 2 years, I’d be recommending 120TB NVME drives backed by HDD’s.
For now, we are limited to 60TB per NVME and we only deploy those in wicked installations.
1 Like
Asus ProArt X670E-CREATOR WIFI
AMD Ryzen 9600X/9700X/7700
2x Broadcom HBA 9500-16i
4x Micron MTC20C2085S1EC48BR
(Micron 32GB DDR5-4800 ECC UDIMM 2Rx8 CL40 | MTC20C2085S1EC48BR | Crucial EU)
1 or 2x Crucial T500 4TB (cache)
1x Crucial T500 1TB (boot)
1x Intel E810 NIC and plug into the last 16x (4x) slot
A bunch of Toshiba Enterprise HDDs? 
Something like that?
3 Likes
That would do, I was wondering if I should switch to epyc/Xeon for this build, and a server mobo with ECC?
1 Like
ECC is supported on the motherboard
2 Likes
It is supported, but not a good recommendation for prod server hosting such a large pool.
Ideally you would be running epyc or xeon.
Yep, I noticed that it is supported, but the mobo won’t be able to take 256gb or even 128gb, looking through their supported mem lists.
We have several users on the forums using 128Gb of RAM including myself (I’m using that exact model) but a 7900 (non X) CPU instead.
2 Likes
Again, it’ll work but not recommended for prod NAS at this level
Based on what? We have no idea about budget etc so your claims are at this point just pointless.
mission critical
mission critical
not a homelab
production environment
I believe he wants enterprise grade recommendations
1 Like
Well, in that case you should’ve rather recommend a failover solution from the start.
…and you seem to have missed “Storage will mainly be used for video backups.”
1 Like
This is for a small business, and though the current setup was developed out of neccessity and homebrew, i am looking to have an enterprise level solution in place. The budget will be dictated by the need. I mean, we are going to be spending 10k Canadian peso’s on hard drives alone, so having a solution that will be stable and reliable are top concerns.
1 Like
absolutely
and my fault
Given that AMD “rebrands” AM5 CPUs is Epyc I don’t see any reason to dismiss it all.
@TryTwiceMedia
You have also yet to suggest something in regards of actual hardware
because Ryzen platform “is not stable”? AMD thinks otherwise, check Epyc 4004 which is a relabeled Ryzen. There are actual commercial data center products around Ryzen.
And @diizzy put up a magnificent list of components. This is a better server than most OEM products
2 Likes