HI Alexander, Thanks for the update. Initially, I also thought of deploying Ceph but ceph is quite difficult to set-up and manage. Moreover, it's also hardware demanding. I think it's most suitable for a very large set-up with hundreds of clients. What do you think of MooseFS ? Have you or anyone else tried MooseFS. If yes, how was its performance? Thanks, Gagan On Thu, Apr 17, 2025 at 1:45?PM Alexander Schreiber <als at thangorodrim.ch> wrote:> On Thu, Apr 17, 2025 at 09:40:08AM +0530, gagan tiwari wrote: > > Hi Guys, > > We have been using OpenZFS in our HPC environment for > > quite some time. And OpenZFS was going fine. > > > > But we are now running into scalability issues since OpenZFS can't be > > scaled out. > > Since ZFS is a local FS, you are essentially limited to how much storage > you can stick into one machine, yes. > > > So, I am planning to use Gluster on top of OpenZFS. > > I don't think that is giving you the kind of long term scalability > you might expect. > > > So, I wanted to know if anyone has tried it. if yes, how it was and any > > deployment guide for it. > > I'm running GlusterFS in a small cluster for backup storage. > > > We have an HPC environment . Data security and extremely fast read > > performance is very important for us. > > > > So, please advise. > > For that use case I would actually recommend Ceph over GlusterFS, since > that can be pretty easily scaled out to very large setups, e.g. CERN is > using multiple Ceph clusters sized at several PB and their use cases > usually include very fast I/O. > > Another concern is that Ceph is being quite actively developed whereas > GlusterFS development seems to have slowed down to ... not much, these > days. > > Kind regards, > Alex. > -- > "Opportunity is missed by most people because it is dressed in overalls and > looks like work." -- Thomas A. Edison >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20250417/19c605c8/attachment.html>
On Thu, 2025-04-17 at 14:44 +0530, gagan tiwari wrote:> HI?Alexander, > ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Thanks for the update. Initially, I? > also thought of deploying Ceph but ceph is quite difficult to set-up > and manage. Moreover, it's also??hardware demanding.You are of course entitled to your own opinion but I'd like to point out that ZFS+Gluster carries a lot of considerations and foot-guns of their own. Saying that Ceph is hardware demanding is also quite misleading as you can install a Ceph cluster on VM's or RPi's with 2 GB of RAM but that's hardly the case for any HPC environment so the argument kind of falls flat. Personally I'd argue that having two intermingling systems is more complex than having one (bigger) system to learn and manage. Now, is Ceph perfect? No. But is it the most consistent and well-documented in all aspects? Also no :) Is it the safest choice of the two? Definitely yes! /K> I think it's?most suitable for a very large set-up? with hundreds of > clients.? > > What do you think of MooseFS ?? Have you or anyone else tried > MooseFS.?If yes, how was its performance? > > Thanks, > Gagan > > > > On Thu, Apr 17, 2025 at 1:45?PM Alexander Schreiber > <als at thangorodrim.ch> wrote: > > On Thu, Apr 17, 2025 at 09:40:08AM +0530, gagan tiwari wrote: > > > Hi Guys, > > > ? ? ? ? ? ? ? ? ? We have been? using OpenZFS in our HPC > > > environment for > > > quite some time. And OpenZFS was going fine. > > > > > > But we are now running into scalability issues since OpenZFS > > > can't be > > > scaled out. > > > > Since ZFS is a local FS, you are essentially limited to how much > > storage > > you can stick into one machine, yes. > > > > > So, I am planning to use Gluster on top of OpenZFS. > > > > I don't think that is giving you the kind of long term scalability > > you might expect. > > > > > So, I wanted to know if anyone has tried it. if yes, how it was > > > and any > > > deployment guide for it. > > > > I'm running GlusterFS in a small cluster for backup storage. > > > > > We have an HPC environment . Data security and extremely fast > > > read > > > performance is very important for us. > > > > > > So, please advise. > > > > For that use case I would actually recommend Ceph over GlusterFS, > > since > > that can be pretty easily scaled out to very large setups, e.g. > > CERN is > > using multiple Ceph clusters sized at several PB and their use > > cases > > usually include very fast I/O. > > > > Another concern is that Ceph is being quite actively developed > > whereas > > GlusterFS development seems to have slowed down to ... not much, > > these days. > > > > Kind regards, > > ? ? ? ? ? ? Alex. > > ________ > > > > > > > > Community Meeting Calendar: > > > > Schedule - > > Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC > > Bridge: https://meet.google.com/cpu-eiue-hvk > > Gluster-users mailing list > > Gluster-users at gluster.org > > https://lists.gluster.org/mailman/listinfo/gluster-users-------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 659 bytes Desc: This is a digitally signed message part URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20250417/6754bd3c/attachment.sig>
Gagan: Throwing my $0.02 in -- It depends on the system environment of how you are planning on deploying Gluster (and/or Ceph). I have Ceph running on my three node HA Proxmox cluster using three OASLOA Mini PCs that only has the Intel N95 Processor (4-core/4-thread) with 16 GB of RAM and a cheap Microcenter store brand 512 GB NVMe M.2 2230 SSD and my Ceph cluster has been running without any issues. As someone else mentioned, to state or claim that Ceph is "hardware demanding" isn't wholly accurate. As for management, you can install the ceph-mgr-dashboard package (there is a video that apalrd's adventures put together on YouTube which goes over the installation process for this package, if you're running Debian and/or Proxmox (which runs on top of Debian anyways).) From there, you can use said Ceph manager dashboard to do everything else, so that you don't have to deploy Ceph via the CLI. I was able to create my erasure coded CRUSH rules using the dashboard, and then create my RBD pool and also my CephFS pool as well. (The metadata needs a replicate CRUSH rule, but the data itself can use erasure coded CRUSH rule.) If your environment is such that you can do this, then Ceph might be a better option for you. If you look at the benchmarks that tech YouTuber ElectronicsWizardry ran, ZFS is actually not all that performant. But what ZFS is good for are some of the other features like snapshots, replications, and it's copy-on-write schema for modifying files (which again, based on the testing that ElectronicsWizardry ran, does indeed create a write amplification effect as a result of the copy-on-write architecture) Conversely, if you're looking for reliability, the more nodes that you have in the Ceph cluster, the more reliable and resilient to failures the Ceph backend will be. Thanks. Sincerely, Ewen ________________________________ From: Gluster-users <gluster-users-bounces at gluster.org> on behalf of gagan tiwari <gagan.tiwari at mathisys-india.com> Sent: April 17, 2025 2:14 AM To: Alexander Schreiber <als at thangorodrim.ch> Cc: gluster-users at gluster.org <gluster-users at gluster.org> Subject: Re: [Gluster-users] Gluster with ZFS HI Alexander, Thanks for the update. Initially, I also thought of deploying Ceph but ceph is quite difficult to set-up and manage. Moreover, it's also hardware demanding. I think it's most suitable for a very large set-up with hundreds of clients. What do you think of MooseFS ? Have you or anyone else tried MooseFS. If yes, how was its performance? Thanks, Gagan On Thu, Apr 17, 2025 at 1:45?PM Alexander Schreiber <als at thangorodrim.ch<mailto:als at thangorodrim.ch>> wrote: On Thu, Apr 17, 2025 at 09:40:08AM +0530, gagan tiwari wrote:> Hi Guys, > We have been using OpenZFS in our HPC environment for > quite some time. And OpenZFS was going fine. > > But we are now running into scalability issues since OpenZFS can't be > scaled out.Since ZFS is a local FS, you are essentially limited to how much storage you can stick into one machine, yes.> So, I am planning to use Gluster on top of OpenZFS.I don't think that is giving you the kind of long term scalability you might expect.> So, I wanted to know if anyone has tried it. if yes, how it was and any > deployment guide for it.I'm running GlusterFS in a small cluster for backup storage.> We have an HPC environment . Data security and extremely fast read > performance is very important for us. > > So, please advise.For that use case I would actually recommend Ceph over GlusterFS, since that can be pretty easily scaled out to very large setups, e.g. CERN is using multiple Ceph clusters sized at several PB and their use cases usually include very fast I/O. Another concern is that Ceph is being quite actively developed whereas GlusterFS development seems to have slowed down to ... not much, these days. Kind regards, Alex. -- "Opportunity is missed by most people because it is dressed in overalls and looks like work." -- Thomas A. Edison -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20250417/29a7b372/attachment.html>
On Thu, Apr 17, 2025 at 02:44:28PM +0530, gagan tiwari wrote:> HI Alexander, > Thanks for the update. Initially, I also > thought of deploying Ceph but ceph is quite difficult to set-up and manage. > Moreover, it's also hardware demanding. I think it's most suitable for a > very large set-up with hundreds of clients.I strongly disagree. I run a small (3 nodes) Ceph cluster in my homelab and following the official docs it's pretty easy to set up. The hardware demands mostly depend on what performance one needs - the more performance (e.g. NVMe storage and 100 GBit networking) one wants, the more powerful hardware one has to provide, as usual. My nodes are Intel D-1521 with 64G of (ECC, of course) RAM and Connect-X 4 cards running at 10 GBit and storage on HDDs which provide reasonable performance for my needs - not a HPC setup, of course.> What do you think of MooseFS ? Have you or anyone else tried MooseFS. If > yes, how was its performance?Last time I looked, MooseFS needs a commercial license for the full feature set (e.g. highly available metadata (_that_ is not negotiable!), erasure oding, Windows clients) which killed it for my non-commercial use case. Kind regards, Alex. -- "Opportunity is missed by most people because it is dressed in overalls and looks like work." -- Thomas A. Edison
Lets back up a bit. What are you using zfs for? The use case will defiantly effect your choice of solution. On 2025-04-17 05:14, gagan tiwari wrote:> HI?Alexander, > ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Thanks for the update. Initially, I? > also thought of deploying Ceph but ceph is quite difficult to set-up > and manage. Moreover, it's also??hardware demanding. I think it's?most > suitable for a very large set-up with hundreds of clients. > > What do you think of MooseFS ?? Have you or anyone else tried > MooseFS.?If yes, how was its performance? > > Thanks, > Gagan > > > > On Thu, Apr 17, 2025 at 1:45?PM Alexander Schreiber > <als at thangorodrim.ch> wrote: > > On Thu, Apr 17, 2025 at 09:40:08AM +0530, gagan tiwari wrote: > > Hi Guys, > >? ? ? ? ? ? ? ? ? We have been? using OpenZFS in our HPC > environment for > > quite some time. And OpenZFS was going fine. > > > > But we are now running into scalability issues since OpenZFS > can't be > > scaled out. > > Since ZFS is a local FS, you are essentially limited to how much > storage > you can stick into one machine, yes. > > > So, I am planning to use Gluster on top of OpenZFS. > > I don't think that is giving you the kind of long term scalability > you might expect. > > > So, I wanted to know if anyone has tried it. if yes, how it was > and any > > deployment guide for it. > > I'm running GlusterFS in a small cluster for backup storage. > > > We have an HPC environment . Data security and extremely fast read > > performance is very important for us. > > > > So, please advise. > > For that use case I would actually recommend Ceph over GlusterFS, > since > that can be pretty easily scaled out to very large setups, e.g. > CERN is > using multiple Ceph clusters sized at several PB and their use cases > usually include very fast I/O. > > Another concern is that Ceph is being quite actively developed whereas > GlusterFS development seems to have slowed down to ... not much, > these days. > > Kind regards, > ? ? ? ? ? ? Alex. > -- > "Opportunity is missed by most people because it is dressed in > overalls and > ?looks like work."? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? -- Thomas > A. Edison > > > ________ > > > > Community Meeting Calendar: > > Schedule - > Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC > Bridge:https://meet.google.com/cpu-eiue-hvk > Gluster-users mailing list > Gluster-users at gluster.org > https://lists.gluster.org/mailman/listinfo/gluster-users-- Alvin Starr || land: (647)478-6285 Netvel Inc. || home: (905)513-7688 alvin at netvel.net || -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20250417/92c4a13f/attachment.html>