Hey friends, looking for advice on how I can possibly get the best of both worlds out of this situation…
Some Background:
I’ve been running a Plex server for about 2 years now. 2018 Mac Mini, DAS: Pegasus R4 with 4x 12TB drives configured in RAID5. It’s worked great, no complaints. I’m nearing the 36TB capacity of my array and have been researching the best course to upgrade for a couple months. Very interested in the idea of building a proper server and moving to Proxmox so I can start doing some other homelab stuff with the machine as well.
Yesterday, I got a new job which starts in February, so I have a little bit of time to figure things out.
For the new job, I will need to build a RAID with 70-100TB of usable space. I’ll be offloading 1TB and 2TB NVMe’s to 2 destinations. Destinations are the HDD RAID I’m building as well as to a 4TB NVMe external drive. Everything must be run through an xxhash64 checksum. One downside, is I’m required to use MacOs for the job. The offload has to be pretty fast, without breaking the bank.
The cheapest/easiest solution I’ve come up with is:
- Buy Mac Studio M1 Max ($1,579 refurbished from Apple)
- Buy OWC Thunderbay 8 enclosure ($899.99)
-Fill the Thunderbay with 20TB IronWolf Pro drives ($1800 for 6x 20TB)
Pros to this setup: Relatively cheap given the parameters. No building anything or relying on my (lack-of) skill in building a DAS/NAS. Good warranties on the computer and enclosure if something goes wrong and needs replaced fast.
Downsides to this setup: I really don’t like the idea of using Softraid. I would much prefer a hardware raid. The Thunderbay is big and power hungry, and I’d prefer something rack mounted.
I considered the idea of building a NAS…But then to move files from the source drives to the NAS, would I just need a 10Gb switch to establish a local network on my work truck? (I’ve never done this, only used DAS for work) I also have no idea what kind of read/write speeds I could expect from a NAS spinning the IronWolf Pros.
What I’m looking for is advice on possibly building a DAS or NAS that I could use for the job and, when the job is over, reuse as much of it as possible in my new Plex server / Homelab build. Ideally, it would be compatible with both MacOS and Linux, have hardware RAID control, minimum 8 bays, be rack mounted and power efficient.
Based on the price of the above setup, I’ll say my budget is $2,500 excluding the cost of HDDs.
Things I already own that may or may not be helpful: 2018 Mac Mini i7 w/ 16GB DDR4 RAM, Pegasus R4 with 4x 14TB IronWolf Pro Drives
Thanks for your time and I appreciate any feedback you can give!
You must log in or # to comment.
Other commenters have good advice re software RAID. I’ll just reiterate you should investigate that. TrueNAS is what I use, and it’s great.
I’ll add some thoughts about the application. Is this for professional work, as in, someone is going to be counting on you to forever more have the files you mention available? If so; you have basically just one copy (I think) of the info, that’s what gets backed up in the NAS/RAID. if people are needing this data to be safe and protected against all manner of things that happen to data, you’ll need to consider another backup option, offsite, like a cloud provider. Something like Backblaze, AWS, or Google or something. Another great feature of software RAID solutions like truenas or unsaid is that they make it easy to setup rsync or something similar to automate these cloud (or local) backups for you.
If this is just you trying to make your workflow more convenient, that’s a bit different, but still know you’re a little light on the backup side of things. No matter what RAID type you use, it’s not a backup.
It’s a whole world, especially with fast transfer speeds. I recommend you seek paid advice
Need a lot more context such as types of files you are transferring. Is it video files, or word docs
checksums are typically calculated client side. The server will do its own thing re raid/parity.
Eg if you were doing video files I’d recommend Hedge for Mac. If you are windows I’d recommend Terracopy Pro but avoid on Mac
You’ll want a separate network, 10G or more for fast transfer speeds, ofc speeds are limited by your raid. How many people need to be connected to it?
If just you can have a separate 10G network card on your server, and then an external Sonnet 10G Thunderbolt 3 adapter for your MacBook or however you are connected to the serve
If more then you need a network switch
I wouldn’t use a Mac as a host with drivers I would build a separate Linux system for your server with the drives connected. You can then share the storage pools over the network via smb or nfs
Truenas is a good start if you have time to tinker
If you want something really easy I’m selling a Synology disk shelf and expansion, holds 12 disks each, super easy gui
Thanks for the response and all the information! Sorry for not being clear about several things. I’ll try to clarify.
I will be offloading Arri Alexa 35 Capture Drives (1TB and/or 2TB) to a “Shuttle drive” as well as a “master backup”. The capture drives (source) are NVMe. The “Shuttle drive” (destination 1) is NVMe, the “master backup” (destination 2) is the HDD RAID I would like to build.
It’s all video files. Specifically, 4K ProRes 4444XQ, LogC4 files in .mxf’s.
For checksums, I meant that I will be generating the xxhash64 checksums from the codex compact drives to the 4TB NVMe shuttle drives and my RAID.
For checksum and transfer software, I’ve used Silverstack and ShotPutPro in the past.
I’ve always avoided Hedge because of what an assistant editor told me years ago (and this part is WELL beyond my understanding, so please correct me if I’m way out of line) to paraphrase what I was told:
“Hedge does not create a legitimate checksum. Hedge begins the checksum process while files are still moving between the RAM and the destination drive. A legitimate checksum must be between the destination and the source. While unlikely, there is a possibility that information could be lost between the RAM and destination making the checksum Hedge performs illegitimate.”
This is what I was told right when Hedge first came out and anytime I’ve seen Hedge asked about it since, they’ve ignored the question or skated around it. If this is incorrect, I would love to know. Hedge is incredibly affordable and faster than ShotPutPro. I just always assumed it was faster because they were cheating the checksum (based on the info above).
Back to your questions/variables.
I will be the only person that needs to access the “master backup”/RAID.
Based on what you said, I think my best bet is to build my Linux based server with a 10G network card and then network directly into my 2018 mac mini. At that point, the bottleneck is the 260-285MB/s speed of my 7200 RPM drives, right? Is there any reason why my 2018 i7 6-core with 16GB of DDR4 would slow the process down?
Originally, I thought I was going to be transcoding ArriRaw to HDE files, so I was going to get the Mac Studio for the 24-core GPU…but now that there’s no transcode I’m hoping maybe my 2018 mac mini will be sufficient?
Those 6 7200RPM drives will be able to saturate your 1Gbps link and will even give your 10 gig link a pretty decent workout. Remember, when working with RAID, you’re dealing with multiple IO, so you can take throughput of one of those drives and multiply it by six, and that’s effectively what you’re going have available. I would personally recommend using a good set of SAS drives, over SATA, to maximize your bandwidth server side.
Thanks for commenting! I’m not sure if I can afford to go with 6x 20TB SAS drives over the 6x 20TB SATA drives, but will have to consider that…
Remember, when working with RAID, you’re dealing with multiple IO, so you can take throughput of one of those drives and multiply it by six, and that’s effectively what you’re going have available.
Can you explain this for me or point me to a resource to learn what you mean? My knowledge of RAID is limited to how many drive failures it can handle and the boosts in read/write speeds. I will be using RAID 5, possibly RAID 6 for my situation, so no write speed gains either way.
Let me know if I’m missing something here. My single mac computer will be the only client able to access the server.
Where I need the most speed is writing from the NVMe source to the RAID. The only time I’d be reading the data on the RAID is when it’s generating the checksum after the data transfer is complete, right?
I don’t need to access any of the files kept on the RAID while I’m offloading the footage. So the write speed from the source to the RAID should be 260-285Mbps, I think?
He is wrong about Hedge, it’s fantastic software. Copied petabytes through it without error
Hedge does source verification which is legit, it checksums in ram, but so does all xx64 that’s why faster ram helps re your question about using a Mac mini
The chance of a non ecc ram error is incredibly low. And worthit for the speed advantage, you don’t need to generate another destination checksum afterwards, it does it at the same time so won’t reallllyyy be reading from the array.
If you want pure speed I would get a large nvme cache. 2x 4TB ssds in raid 0 means it can hold 2/3 large arri mags before it starts to transfer to the disks
sent from mobile
Can you share more about the requirements and uses? Why does it have to be macOS based…especially if it sounds like you’re keeping the hardware at the end? Is it because these NVMes are APFS?
If I wanted a Linux server long term but needed it to be a Mac for a temporary job I’d build an Proxmox server, spec it accordingly, and run a macOS VM. Not the best long term solution but you can get it running and disable updates so it is stable for shorter timespans.
Mac user here. Similar setup different capacities similar models.
I have an M1 a 2018 and a Mpro 2013.
2 thunderbay4 type enclosures (Akitio thunder 3) on the M1
1 Thunderbay4 TBT2 and a Lacie Duo raid dock on the 2013 Mpro.
I use the 2018 with its 4 port TBT3 ports as a 40GB switch between them plus as VM machine. This also gives me 2 more TBT3 port for networking or expansion.
If you can afford it use soft raid. The other good alternative is using ZFS. I currently use it.
If you want more details let me know.
software raid has come a long way, and it’s worth considering for your setup. it offers flexibility, maintenance tools, and integration with virtualization platforms like Proxmox. plus, it’s more cost-effective than a hardware raid setup. consider exploring software raid options before making a decision.
I have synology and was considering switching to truenas and the latter is PRICEY in comparison. QNAP has had multiple issues with breaches and ransomware so I’d avoid. Synology has been pretty good in that regard.
I’m trying to Hannah Montana this
Huh?
A Disney show from the 2000s which chronicled the attempts of a teenage pop-star (stage name Hannah Montana) to maintain a secret identity in order to have the “best of both worlds”: the many benefits of fame, and the privacy and normalcy of relative anonymity.
One thing for you to consider: Software RAID is no longer a dirty word and somewhat suspicious.
As current processors have a lot of power available, it becomes unnecessary to have a controller and a specialized processor. You can find good and solid RAID systems in free software, much more tested than a controller from… a beginner manufacturer.
More flexible configurations, more maintenance and recovery tools, more sophisticated security, better logs and, especially, great integration into virtualization platforms like Vmware and Proxmox.
Just going to second this. Software raid is fine. I have hardware raid cards that I use as basic HBA’s to attach more drives which I then software raid rather than use the hardware raid functionality. In a decent number of cases software raid can be preferable. 10 to 20 years ago this was not the case.
UNRAID FTW!
