That being said - I get the temptation of dissatisfaction with some products.
But I come back to DIY NAS is best for people/shops with:
In house IT staff (FT or contract)
Are the IT staff for their shop as part of their job
More time than money - for variety of scenarios
OCD about their tools and processes
Like to have a community recognition for building custom systems
Crazy unique requirements or scale
I’m a small boutique shop. And yes, I’m my own IT department, but I’d rather be spending time on billable time than fixing my infrastructure, and my weekend either with family or learning skills that I can bill for.
Most people I know are more time constrained than cost constrained.
And when things go sideways while on a client deadline, it’s nice to file a support ticket and get some help rather than frantically searching Google in the hope of getting back on schedule soon. I’m happy to look past some product shortcomings in exchange for that.
There’s a place for them, but neither DIY nor system integrator, nor OEMs are de-facto solution for everyone scenario.
I don’t know their storage solutions, but they’ve built 3 of my workstations and they are generally very well regarded for the solutions they configure/build.
I looked into Syncthing for my synology so it syncs with my server… My ubuntu has a hard time with dropbox that i have to use with one of my clients. Synology NAS is great with dropbox but Syncthing is not very well supported on the NAS and requires some work to get it on there. I wish Rsync or something part of the NAS software that syncs continuously.
i have decided to just suck up the costs and go full lucid link, less admin overhead and it means i can save $20K right now in not having to buy a nvme nas with 100/25Gbit and all that.
Fixes nuke speed issues too, only downside is that they dont have a studio caching server, that would be amazing.
LOVE the idea of lucidlink. One day to remove the admin side of server filesharing. Its a pain but its also kind of interesting! theres a little joy when things copy well.
Yea the problem is what i want is a “lucid cache” not a fileserver with a mounted lucid partition.
Like i still want everyone in the office to conenct using the lucid client and do local caching but also pull data from a big lucid “aggregator” so i dont have to be WAN bottlenecked.
Wan in germany is Extremely expensive, our 1G/1G line is around 1000€/Month, so 2-10Gbit wan lines are compeltely out of the question.
Sure. I guess my point is that if you pin the cache on an NFS server and have your clients connected to that you’re never caching a file more than once. Sure, the clients won’t have the same speed of access… they’ll bottom out at the speed of the NFS/Cifs mount but at least you won’t be hammering your WAN anymore than you have to.
It would be dope though to have a clustered approach to caching though–almost bit-torrent style or something.
yea dropbox has that for example, one big adavantage with lucid for me is just that everyone has the data on fast storage, what we do is very ansynchronous , LOTS of reads , very few writes compared.
Now with 4.6K beign standard here, 10Gbit really isnt enough both for resolve and Flame, but also nuke…
So my choice was, go full 100Gbit NVME server, put fibre everywhere or come up with a scheme were we scale down project sizes by a lot, ill still keep my 10gbit synology setup (3-2-1 all the things) as a cold storage, where i pull projects from lucid (hot) as soon as its done.
Nuke has a localization feature which kinda made this into less of an issue, but writing with nuke to the nas was still CRAZY slow.
This is all fixed now, i just threw 4Tb NVMEs into every machine, they are cheap and give me speeds not even possible with networking unless your name is alan latteri.
also because we are hybrid , this has a lot of benefits, i have also made a cloud VPS with a 2.5Gbit port that mounts lucid with a simple filebrowser web interface so now i can share files like I can with wetransfer/dropbox directly from lucid took me like a night of coding.
@finnjaeger in your LL workflow, are you working in Flame completely unmanaged? What does it look like if you have to swap a job to another artist at delivery time?
i always work unmanaged, full publish workflow and everything, best case I just swap aritst, worst case I swap machines and have to re-download the dwab compressed EXRs that ive rendered, not a big deal at all even at 1Gbit
You can get 25gig NICs off eBay for less than $100.
You can get super cheap fiber from FiberStore.
For probably less than $4k you can outfit your studio to be able to play in realtime just about any format, and not have to worry about any of the above discussions.
We’ve got a couple of FS switches from there and have had no issues. 25GbE infrastructure here too but playing off a PixStor solution. The 25GbE NICs have come way down in price too, we’re using ATTO I believe but there are lots of cheaper options.
I spent a bit of time looking into ways for remote artists to work with a shared namespace on local storage but didn’t quite find anything that was quite right. LucidLink was the most promising but the benefits did not outweigh the cost unfortunately.
There is some amazing distributed file system solutions out there, such as QuoByte, but that doesn’t quite suit your use case either. Could be worth having a chat to those guys as they did have some interesting tools they were working on and it would be BYO Hardware which keeps costs down.
I just go straight RockyLinux, Hardware Raid and XFS. The array is only limited by the PCI bus speed and connects to our network at 100gigE.
In our Flame S+W server we have about 170TB of SSD storage.
On our general NAS, we have ~700TB of hard drive.
This stuff is all 5+ years old now. It could all be built cheaper and all fit within 8u or less.
Unless you wanted to have node level redundancy, or 10s of petabytes, I don’t see the need for a distributed filesystem. Its more complex, costly, and require much more hardware.
Doing construction tomoay down fibre everywhere = $$$
regardless of that - none of this would have solved my initial issue which is hybrid work and scaling up/down , this does the trick, people can open files on their local machines if they want to while beign able to remote in , all of that.
Also a single nvme still beats 25G by like a lot… especially with nuke and nukestudio read/writes, anything network is garbage.
I also never have to use flame caching anymore which is niiiice.
The way i am making lucid work is to really do some housekeeping with files, we always drag around massive amounts of data that nobody needs/wants , this starts with flame archives which are a complete waste of life.
Goes on with DWAB exr renders and trimming the OCF before it even hits lucid, so far i am 3 commercial projects in and I have barely hit 1TB, which makes all of this rather affordable. All of this plus smart lifecycle management is where its at for me, stuff is very slim and fast without lacking quality.
caching in flame is just re-encoding source codecs to whatever you set flame to do and saving that on your framestore/fast storage, my stuff via lucid is there allready and there isnt a case where anyone would load arriraw rushes into flame, it all goes through ingest and gets converted to EXRs, its all VERY VERY tightly pipelined to enable those workflows.
So no flame generated ridicolous media, no archives, all is published and works via openclips. Nuke and Houdini are unproblematic as well.
Oh and additonally I can just use paperspace/ aws machines to scale the studio in a matter of minutes by just mounting the lucid drive, its very convenient. producers at home love it too, they can just space-bar play the Dailies , easy.
I could probably do a cheap 25Gbit setup with just like 16Tb of NVMEs, I could, but with Lucid i can just click a button and onboard a freelance colorist in country X .
We are also thinking of setting up a color suite in a different city which this would become genius for that as well.
especially with nuke and nukestudio read/writes, anything network is garbage.