Out of interest to the community would iX share the percentage of their Enterprise customers that currently use CORE vs SCALE?
Feel free to like this post if you’d be interested in an answer.
5 Likes
Its a useful question.
For the Community… SCALE is about the same number of users as CORE 13.0. It will be larger in Q3.
For the Enterprise: Most mature product users are on 13.x still. X-Series, M-Series. We are only recommending migration for these if there are specific features needed. This will change with 24.4.2 and 24.04.3 There are now equivalent stability, extra performance and extra security. To make this switch, there are a few minutes of downtime, so we don’t make the switch without planning it.
For new products: F-Series, H-Series, they are all on the 24.04 version.
So, in terms of new deployments for Enterprise customers it will be about 50/50 in 2H 2024.
I’d expect the crossover in terms of total volume of Enterprise users in 2025.
3 Likes
I should add that TrueNAS 13.3 will be an update from 13.0 that doesn’t require any planned downtime for HA systems. Failover from one controller to another can be done while keeping VMs and other applications working. There is just a stall in I/O during the failover.
For conservative or mission-critical users this 13.3 update may take them well into 2026 and beyond. This update minimizes risk and downtime and requires no retraining for the existing WebUI. It still gets the benefits of OpenZFS 2.2. So, we don’t expect a rapid migration to SCALE.
2 Likes
Network: FreeBSD consistently out preforms Linux (Debian) without tuning (with tuning Linux can typically match). Are you guys preforming those tunings to get the same network performance on SCALE for the enterprise users? Also, were you making network tunings to COREs kernel? I am on the thought-track of “FreeBSD is typically the choice in low-latency, high consistency, high secure network applications”; are there benchmarks given to enterprise customers showing kernel tunings to match CORE? And/or are they giving any feedback about any benchmarks they’ve done? Generally speaking, of course.
Yes, of course we tune for each OS and platform. It takes time to test and get it right.
Dragonfish is significantly outperforming 13.0, but we also expect 13.3 to be improved.
Blog here: TrueNAS Dragonfish Performance Breathes Fire
This chart shows the difference with our M50 platform:
1 Like
Nice but that comparison is a little one-sided don’t you think? I mean your testing older against newer and newest (discrepancies are expected). What I take is that my very much older FreeBSD can hold its own against the newest Linux Kernels (-i.e. shouldn’t it be 13 v Bluefin?). At any rate, thanks.
Its Current CORE vs SCALE… as indicated, we expect CORE 13.3 to be a bit faster.
However, most of the performance improvements seem to be in the OpenZFS and protocol stacks… not the underlying OS.
Hi, I reply on this topic to get some advices because today I’m a bit lost.
Straight to the point, I had a great infra using FreeBSD jails on truenas core and I blindly migrated to scale, so I’ve lost most of my services. Didn’t had the time to convert them in Scale containers and to be honest, I wasn’t very enthousiastic. IX containers like Plex and Nextcloud are slow to come up, when they work.
Several months after, I’ve migrated my pool on wider disks and Scale had provided a very frendly and well design interface. Few weeks after, I’m migrating again on Truenas Core on a version sharing the same ZFS version and … Disappointing. Truenas Core has not been improved, not even on the graphical user interface. Where changing disks was very easy with Scale, just finding the way to do it on Core was more difficult.
So I’m thinking about going on Scale in all conscience this time but before, I’d like to have some feedback on comparaisons in general. I had once time to have hard work on making things work perfectly and I’m today running after it. Does Scale can be today a Proxmox alternative in managing custom containers ? Maybe someone could convince me to go back on Scale and let Core behind once and for all … Or the opposite. Please help me get a clearer vision, my home server is actually stopped and I don’t know which way to go.
I don’t have any benchmarks for TrueNAS specifically, but Netflix has plenty of numbers on FreeBSD vs Linux for their streaming servers and FreeBSD has significantly more throughput over Linux when you’re pushing 400 Gbps of data all the time, which is the reason they prefer it over Linux.
However, it is worth noting that they do NOT run the same code as us mortals. They track FreeBSD-CURRENT and actively contribute back to upstream.
Further reading if you care to see more geeky stuff:
I’m actually interested in this question also. Not so much for the Apps/containers (I prefer jails for containers), more for the VM’s since that’s mainly what I’m running Proxmox for.
Almost all development resources are going to Scale. Go Scale. Upcoming Scale version nicknamed Fangtooth is supposed to be a direct upgrade from Core.
1 Like
Well, given how iXsystems acted on their promises in the past, I will ignore any forward-looking statements from side for now.
While I like FreeNAS/TrueNAS as products, the company has lost my trust for the time being. It will take a few years to regain it and I do hope that’ll happen.
4 Likes
I wonder if they ever got their delivery?
CURRENT?! That’s crazy (that takes some serious skill)!
I understood maybe 4 words in that presentation, BTW.
It’s nice to hear this coming from someone else every now and then.
2 Likes
Hello, long time user and FreeNAS Mini owner currently running TrueNAS CORE 13.0 U6.2. I registered a forum account to add my voice to this topic.
Given Netflix’s continued high performance use case with FreeBSD and the issues with Linux quality and Network Stack, it would seem to me that the advantage of more immediate hardware support in exchange for lower quality software is not worth the trade. The Cathedral vs the Bazaar.
Further, the burden of support for Linux as an administrator may be enjoyable when you are okay with re-learning the “new” way every time you upgrade because they are breaking compatibility, deprecating the tools you know (rc init, ifconfig, in exchange for systemd and ip, for example) but less so when you observe that the old tools are sufficient and the “new” way markedly inferior. When I was a child, learning unix, breaking compatibility was no different than simply learning the next cool new tool. Now I rely on these tools as infrastructure. The BSD systems I set up over a decade ago are still running with less than an hour’s administrative maintenance overhead per year for all of them. If I need to add a service, it’s the same thing we did 30 years ago. I spend that daily on my work Ubuntu machines because of its mysterious, unexplainable network stack phenomena like routing to networks when there is no route defined. Docker makes this significantly worse. I rely on networking to be a deterministic activity. Not some fuzzy routing stack.
It appears that even here, we have unresolved network performance degradations on SCALE: “Significantly reduced sync write performance on Scale vs. Core” (not allowed to link).
My question is this, how does the latest SCALE compare to the latest CORE on the FreeNAS Mini hardware in terms of performance? Things like, what are the CPU and RAM load for the same operations? How do I know that the quality is improved? Docker is not a feature that is worth compromise in design quality or performance to me.
I am very happy with the FreeNAS Mini hardware for its intended use case and performance over the years. A big kudos to Jordan Hubbard and the team for bringing us that experience. As I think about it, I can only imagine how difficult it is to accomplish such a feat in software, as the users can just silently use the tool without constantly being in touch due to change dependence. Good technology is transparent. Like TCP/IP. It just works. We barely think about it because what breaks is outside of those things.
Thank you for reading my thought on the matter. Hope y’all are enjoying some holiday time at the end of 2025 here.
Cheers!
great question - i asked a similar one in a different thread.
My main NAS used to be Core until recently when i moved to an all-in-one system (fnOS) that can host VM/containers… I still run a PVE server but am serious considering moving all of my stuff there to fnOS.
I started digging a little more into this and found that Netflix uses UFS, not ZFS, because the overhead (especially of ARC) is not worth it for their use case. Since my use case revolves around file serving, not as much cold storage, perhaps it is worth considering a modern FreeBSD UFS-based approach that aligns more with how Netflix uses it.
The sources on this need verified, but an initial web query resulted in this comparison between UFS and ZFS performance. Netflix achieves 90 Gbps TLS-encrypted serving with ~55% CPU utilization on commodity Xeons using these techniques. TrueNAS with ZFS typically requires 75–85% CPU for equivalent throughput due to ARC (Adaptive Replacement Cache) overhead.
Add to that the memory requirements for ARC and we further reduce the serving capability.
Now I am curious what a modern approach to the NAS architecture might look like with this in mind. Do we have any professionals in the house who can comment? Perhaps UFS at the top layer with the interface to the network and a bottom layer of the lightest weight ZFS for bitrot resistant, self-healing cold storage. Periodic full disk reads on the ZFS to engage the checksum activities. I’ll keep noodling.
Cheers!
with the small data that most of us are managing (<4TB for me), not sure if that’s big of a deal.
I have compared ZFS with straight btrfs + buffers and no difference (over a 1gbps network). so i’m no convinced that i should lose sleep over this.
Also, freebsd comes with ufs vs. zfs versions so you can install samba over freebsd + ufs/zfs and get your own “truenas” equivalent going.
Sure–If you ignore the API, middleware, and UI, which collectively are literally what make TrueNAS, TrueNAS.
If you are running HDDs at that small scale, performance is almost identical. OpenZFS is the determinant of performance… its limited by the HDD IOPS. Most of the improvements in Linux and OpenZFS are for systems which are much larger or all-flash.
There are some difference between Jails, Docker Containers, and VMs but the noticeable differences will depend on the specific Apps.