Stratis is not a new filesystem. It is a set of management tools (written in python) and a daemon (written in rust) that utilizes XFS, device-mapper, and LVM for the base filesystem, and LUKS and Clevis to handle encryption. This is why its performance closely follows XFS. It is just XFS. However, it is thin-provisioned, and automatically grows as needed via the daemon. Stratis is not a complete replacement for ZFS or btrfs. It does not support parity for recovery. Therefore it has no way to recover from a failed block device and must rely on hardware or MD RAID.
The contents of this video were contained within the video. I found the information presented to be very informative. I found the enjoyable parts to be enjoyable. I watched that which I saw. Please don't over analyze this comment. It is to make the semantic analysis algorithm happy
From Straris' own description: " It builds upon elements of the existing storage stack as much as possible. Specifically, Stratis uses device-mapper, LUKS, XFS, and Clevis. ". So I don't think you can really consider this a file system of it's own, and since it uses xfs it certainly doesn't resolve any licensing issues with xfs.
I always figured this was because it was a RHEL specific project. Looking at the website in the past everything was very specifi to RHEL. With that said I've got in running on Archlinux in the past, even as a root system for kicks.
I work with ZFS a lot in my job and Stratis mostly just seems like a poorly done attempt to recreate ZFS. It would be neat if Stratis could have enough adoption to become a solid open source replacement for ZFS that is native to the Linux kernel.
People go on about ZFS licensing forgetting where BTRFS, which is a child’s toy comes from. Also OpenZFS is detached from Oracle as much as Linux itself is disconnected from SCO/IBM. Just use (Open)ZFS.
Exciting... didn't know about this fs before! I just set up my laptops with btrfs. May be when I need to reformat it (2-5 years) Stratis will be a proven tried and tested alternative?
Well I disagree with you on a few points. You just said that stacks of software leads to more bugs.... But that's good encapsulation. LUKS concerns itself with encryption, but ext4 deals with the actual files, journals, stats and so forth. Good encapsulation leads to fewer bugs. Also, you NEED to have different configurations for different use cases. File system matters for databases and mongo databases. Some programs deal with lots of small files, others do better with giant files.
2:00 all those negative claims are false, they only arise in the current unsupervised situation. I am not against refactoring old stuff, but just do it right, when you are doing it and don't just take the easy path back to monolithism.
This is really just XFS + LVM2. I'm not sure what is even the point. Just use BTRFS and be done. Making XFS into. Volume Managing kit is clunky at best. As for resizing, it's XFS which means it cannot shrink, unlike BTRFS which handles shrinking, and very much like ZFS. As for /boot I suspect it would work without much fuss, again it's just XFS. As for automatically growing, that did indeed work.
I would only add, use dm-integrity underneath everything except ZFS and /boot. ZFS includes it already and most boot loaders don’t understand it. For everything else it provides a cheap and effective way to tell if the drive is lying about reading a sector correctly.
I'm wondering if you watched the first half of this video? He spends most of the time explaining his answer to your point. He doesn't like stacks, and wants everything in one system so it can be managed from a single tool. He even said this repeatedly in the course of his presentation. You can disagree with that, of course, but he does make it clear why he holds that view. If you want your alternative view to have any traction you need to present arguments against the perceived need for single tool maintenance. So when you say you are unsure "what is even the point" it says to me that you fell asleep in class
I'm ALSO wondering if you watched the performance tests at the end of the presentation: where BTRFS was much worse than the other three tested systems. So you don't care about FS performance either? Or you were still asleep? Let me guess: you spent time learning BTRFS and don't want to retrain? That is a fair point for some existing users, but it's a point to be balanced against the suggested advantages on offer from Stratis. Existing skill sets are an important issue but not an automatic deal killer. Apologies if my guess is wrong, but I'm then very curious why you are defending the least well performing option?
@@trueriver1950 my apologies, it's unclear to me what you're responding? Never once did my response go into benchmarks, or performance. That said, when my colleagues at Red Hat internally announced Stratosphere, it was actually really slow, but in my own anecdotal testing. XFS however, by itself is really fast, but apparently not with stratosphere daemon orchestrating. Conversely, I've never thought of BTRFS as slow. I think your whole rant is much about nothing.
@@JonDisnard I used to use btrfs, but it was really inefficient on my SSD, with insane write amplification and random slowdowns. And a few times it would get some problem that was practically impossible to fix. I wanted something less buggy, so a few months ago I switched to ext4, and so far it's been just a million times better. (And on rotating disks I'm using ZFS, which has been just perfect.)
Loved the presentation and the easy way you talk about this subject. Very Good! One thing though, my eyes was drawn to the H.A.L screensaver in the background. Doesn anybody know where to fint a copy? :)
Interesitng. Though a couple of things MPIO is below the filesystem, you said ZFS was "bad" to have pools, then said Stratus has pools of storage Good? I would never want a filesystem to assume I want to extend new space to an existing filesystem. Maybe I want a different mount point, or for some other purpose. Also stratus does not appear to be a filesystem itself. Rather it's a volume/pool management of devices over which XFS is layered. From the website Stratis uses device-mapper, LUKS, XFS and Clevis. Stratis may also incorporate additional technologies in the future. So you still need to deal with that layered approach and kernel dependencies. I do agree a formal API is extremely important. That is something all filesystems should have. You mentioned about wanting to know when a drive is failing, XFS only journals metadata. ZFS also does checksuming to detect when a drive fails to properly write or read data. I thought this was interesting about Stratus from the website. It's not a knock just something to note: Stratis tracks information about file systems created using Stratis that XFS is not aware of, and changes made using XFS do not automatically create updates in Stratis. Users must not reformat or reconfigure XFS file systems that are managed by Stratis. Thanks for the video! It's cool to see what is out there now and what's coming down the road. . I agree with you on ZFS / Oracle being a question mark.
Admittedly I stopped watching at 6:34 but as someone familiar with all of the technologies mentioned, I found your arguments for yet another filesystem to be very unsatisfactory.
It is there. Linux supports BeOS. If your distribution of Linux does not have BeOS built-in, you can compile your Linux kernel from source supporting that. Last time I compiled a kernel the version is 5 point something. I believe that the latest kernel still has the BeOS code there.
Zfs is king, but your comments on licensing are a sad reality. and it's always nice to learn about other filesystems. Stratis seems to be the alternative to zfs that I will keep an eye on.
I've given up waiting on BTRFS RAID5/6 for data (although I do use its RAID1 for the system). Watching BCacheFS with interest, but in the meantime OpenZFS has been great for data I care about.
Some of us (me) would like to see a Stratus comparison that also includes OpenZFS, which seems to include pretty much everything in Stratus. It also is easy enough to use for both root an boot.
Bad choice, python that is, fucking syntax changes every release, it's interpreted so it's slow, all the things any sane person doesn't want in a file system.
Is Rust, with its undeniable benefits, yet stable enough for critical usages? I'm not trolling; I don't use Rust, but have the impression that it still is developing and changing more quickly than I would like as the basis for a production grade file system. Ditto Python. And then I read this morning that the Red Hat branch of IBM announced they will stop relea0sing RHEL source, which might include that for the Stratus server and supporting software.
Btrfs had "advantages" not available on other file systems. Benchmark results are perhaps less important. Ext4 had many advances injected into the Linux kernel, published almost daily by The Linux Foundation. The other two file systems might be useful to enterprises handling many terabytes of information per second. Is this what you are discussing - Enterprise users only? Most small business and home users are on Microsoft Windows and-or Apple. Does this type of user interest your video channel? As a Microsoft Windows user, my days is using Microsoft NTFS, with inbuilt Microsoft file compression. Microsoft claims that there is not any benchmark differences. Compression might allow better benchmark results. Btrfs shows bad benchmark results, in your results, Phoronix and other results. For the small business and non professional users, are these differences significant? This new file system is interesting? Why? So many claims and press releases do not state the supposed advantages of the new releases. The outsiders like myself are forced to be fearful of these new products. These "advances" are usually buggy, with many incompatibilities with existing settings. Some products are only alpha and beta releases, but pretend to be ready for use by the general public. Most outsiders therefore will be careful about "Stratis" or whatever new product is released. Is this TH-cam channel another snake oil seller, and how reliable are its claims?
The unbelievable inconveniences folks are willing to go through for "free" and "open" software. At the same time, they don't mind closed and expensive hardware. The open source movement is anti-intellectual at it's core. If something has no physical manifestation, it has no value.
It's all fun ans games until you depend ln someone else to fix stuff, develop the functionality you need or have to live with whatever rules they're pushing down ln you because you own nothing and can't modify anything. You talk as if the FOSS movement came from folk who just wanted to be cool and whatnot, and not because of necessity.
@@andrebrait actually your statement suggest 2 subjects that are not opposed to each other. The bad people who foisted proprietary software (Billy G.?) were all doing it to STOP innovation being freely distributed.
@@andrebrait What unique value or necessity does FOSS afford us that we didn't have before? The best word processors, spreadsheets, CAD programs and operating systems are still closed and commercial. It has neither increased innovation nor lived up to its spurious claims of protecting user freedoms. In fact, FOSS has lead to an even more onerous trend of software as a service and cloud computing where the user's own data and accessibility are locked up behind a paywall. Ironically, it is the large software companies, such as Microsoft, Google and Oracle, who have benefited the most from FOSS. Their smaller less capitalized competitors are forced to not only compete against them, but also against inferior FOSS products. There was once a thriving market for shareware and low cost software products developed and supported by individuals and small companies, but they're no longer viable business models. It becomes no surprise why large companies enthusiastically support FOSS. In addition to solidifying their monopoly, they also get to capitalize on the free labor of eager and passionate FOSS developers under the auspices of FOSS ideals. None of these large software corporations would dare release the source to any of their flagship products, such as Google search or Windows. FOSS has also decimated the opportunities of the once growing middle class of professional software developers. Despite the astounding growth of the computer industry over past decades, the software developer job market hasn't grown at anywhere near the same pace. FOSS enables the technical infrastructure to outsource software development to lower cost global labor markets. Moreover, FOSS lacks the financial barriers such as license, training and support fees of traditional closed software, eliminating a significant advantage of domestic software developers in a global marketplace. In summary, FOSS has been an absolute disaster to software quality, the developer economy, and to the very freedoms it purports to champion and protect.
Funny, your use of "anti-intellectual" seems so narrow. Extremely creative people working in Linux to contribute to so many others, seems like a very good thing.
Stratis is not a new filesystem. It is a set of management tools (written in python) and a daemon (written in rust) that utilizes XFS, device-mapper, and LVM for the base filesystem, and LUKS and Clevis to handle encryption. This is why its performance closely follows XFS. It is just XFS. However, it is thin-provisioned, and automatically grows as needed via the daemon.
Stratis is not a complete replacement for ZFS or btrfs. It does not support parity for recovery. Therefore it has no way to recover from a failed block device and must rely on hardware or MD RAID.
Randomly popped into my feed and while it's over my head, it's very well done and presented! 👏
The contents of this video were contained within the video. I found the information presented to be very informative. I found the enjoyable parts to be enjoyable. I watched that which I saw.
Please don't over analyze this comment. It is to make the semantic analysis algorithm happy
That comment was a comment.
From Straris' own description: " It builds upon elements of the existing storage stack as much as possible. Specifically, Stratis uses device-mapper, LUKS, XFS, and Clevis. ". So I don't think you can really consider this a file system of it's own, and since it uses xfs it certainly doesn't resolve any licensing issues with xfs.
Does XFS have license issues? Wasn't it release as GPL in 2000?
Thank you DJ. Stratis seems like One for the Near Future.
I've wanted to give Stratis a try but I haven't really found any substantial documentation nor videos showing how to use it in any significant way.
I always figured this was because it was a RHEL specific project. Looking at the website in the past everything was very specifi to RHEL. With that said I've got in running on Archlinux in the past, even as a root system for kicks.
I work with ZFS a lot in my job and Stratis mostly just seems like a poorly done attempt to recreate ZFS. It would be neat if Stratis could have enough adoption to become a solid open source replacement for ZFS that is native to the Linux kernel.
Amen to that Sir
The ZFS licensing issue is as much a Linux licensing issue as it is a ZFS issue. There's no problem integrating ZFS into BSD, for example.
Great Video. Lots of Info.
Does anyone know when it will be available in openSUSE Tumbleweed?
I love your channel. Everything you explain is so well done and easy to follow.
Thank you, john
Think I’ll stick with my btrfs config on my laptop, but this is new and interesting. I’ll keep tabs on this new filesystem tho. Thanks for the video!
thanks for the knowledge
No auto use in mainstream distro (Ubuntu/debian), no point trying (as a personal user at least). At least get it to work from gParted?
Covered well in the frrst few minutes of the video.
People go on about ZFS licensing forgetting where BTRFS, which is a child’s toy comes from. Also OpenZFS is detached from Oracle as much as Linux itself is disconnected from SCO/IBM. Just use (Open)ZFS.
Exciting... didn't know about this fs before! I just set up my laptops with btrfs. May be when I need to reformat it (2-5 years) Stratis will be a proven tried and tested alternative?
From what I can tell they are hoping to make this a production release this year or next.
Well I disagree with you on a few points. You just said that stacks of software leads to more bugs.... But that's good encapsulation. LUKS concerns itself with encryption, but ext4 deals with the actual files, journals, stats and so forth.
Good encapsulation leads to fewer bugs.
Also, you NEED to have different configurations for different use cases. File system matters for databases and mongo databases. Some programs deal with lots of small files, others do better with giant files.
What's wrong with EXT4
2:00 all those negative claims are false, they only arise in the current unsupervised situation. I am not against refactoring old stuff, but just do it right, when you are doing it and don't just take the easy path back to monolithism.
This is really just XFS + LVM2. I'm not sure what is even the point. Just use BTRFS and be done. Making XFS into. Volume Managing kit is clunky at best. As for resizing, it's XFS which means it cannot shrink, unlike BTRFS which handles shrinking, and very much like ZFS. As for /boot I suspect it would work without much fuss, again it's just XFS. As for automatically growing, that did indeed work.
I would only add, use dm-integrity underneath everything except ZFS and /boot. ZFS includes it already and most boot loaders don’t understand it. For everything else it provides a cheap and effective way to tell if the drive is lying about reading a sector correctly.
I'm wondering if you watched the first half of this video? He spends most of the time explaining his answer to your point. He doesn't like stacks, and wants everything in one system so it can be managed from a single tool. He even said this repeatedly in the course of his presentation.
You can disagree with that, of course, but he does make it clear why he holds that view. If you want your alternative view to have any traction you need to present arguments against the perceived need for single tool maintenance.
So when you say you are unsure "what is even the point" it says to me that you fell asleep in class
I'm ALSO wondering if you watched the performance tests at the end of the presentation: where BTRFS was much worse than the other three tested systems. So you don't care about FS performance either? Or you were still asleep?
Let me guess: you spent time learning BTRFS and don't want to retrain? That is a fair point for some existing users, but it's a point to be balanced against the suggested advantages on offer from Stratis. Existing skill sets are an important issue but not an automatic deal killer. Apologies if my guess is wrong, but I'm then very curious why you are defending the least well performing option?
@@trueriver1950 my apologies, it's unclear to me what you're responding? Never once did my response go into benchmarks, or performance. That said, when my colleagues at Red Hat internally announced Stratosphere, it was actually really slow, but in my own anecdotal testing. XFS however, by itself is really fast, but apparently not with stratosphere daemon orchestrating. Conversely, I've never thought of BTRFS as slow. I think your whole rant is much about nothing.
@@JonDisnard I used to use btrfs, but it was really inefficient on my SSD, with insane write amplification and random slowdowns. And a few times it would get some problem that was practically impossible to fix. I wanted something less buggy, so a few months ago I switched to ext4, and so far it's been just a million times better. (And on rotating disks I'm using ZFS, which has been just perfect.)
I thought HDF would replace all the other file formats, but it did not.
As in HDF5 for scientific data?
Loved the presentation and the easy way you talk about this subject. Very Good!
One thing though, my eyes was drawn to the H.A.L screensaver in the background. Doesn anybody know where to fint a copy? :)
Interesitng. Though a couple of things MPIO is below the filesystem, you said ZFS was "bad" to have pools, then said Stratus has pools of storage Good? I would never want a filesystem to assume I want to extend new space to an existing filesystem. Maybe I want a different mount point, or for some other purpose. Also stratus does not appear to be a filesystem itself. Rather it's a volume/pool management of devices over which XFS is layered. From the website Stratis uses device-mapper, LUKS, XFS and Clevis. Stratis may also incorporate additional technologies in the future. So you still need to deal with that layered approach and kernel dependencies. I do agree a formal API is extremely important. That is something all filesystems should have. You mentioned about wanting to know when a drive is failing, XFS only journals metadata. ZFS also does checksuming to detect when a drive fails to properly write or read data. I thought this was interesting about Stratus from the website. It's not a knock just something to note: Stratis tracks information about file systems created using Stratis that XFS is not aware of, and changes made using XFS do not automatically create updates in Stratis. Users must not reformat or reconfigure XFS file systems that are managed by Stratis. Thanks for the video! It's cool to see what is out there now and what's coming down the road. . I agree with you on ZFS / Oracle being a question mark.
These videos and this channel is amazing. Will gladly share and recommend.
Admittedly I stopped watching at 6:34 but as someone familiar with all of the technologies mentioned, I found your arguments for yet another filesystem to be very unsatisfactory.
would be nice if you could change filesystem without reinstall
It's possible. You can copy your current system to another storage with other FS, make changes in /etc/fstab, and maybe /boot and that's it.
VFAT?
What I want is an Linux file system that matches the metadata of BeOS. And easy tools to find files.
It is there. Linux supports BeOS. If your distribution of Linux does not have BeOS built-in, you can compile your Linux kernel from source supporting that. Last time I compiled a kernel the version is 5 point something. I believe that the latest kernel still has the BeOS code there.
Zfs is king, but your comments on licensing are a sad reality. and it's always nice to learn about other filesystems. Stratis seems to be the alternative to zfs that I will keep an eye on.
👍!
BTRFS and BcacheFS seem more interesting tbh
I've given up waiting on BTRFS RAID5/6 for data (although I do use its RAID1 for the system). Watching BCacheFS with interest, but in the meantime OpenZFS has been great for data I care about.
I prefer ZFS.
Me too
Some of us (me) would like to see a Stratus comparison that also includes OpenZFS, which seems to include pretty much everything in Stratus. It also is easy enough to use for both root an boot.
What even uses ZFS anymore? OpenZFS is where it's at. Like everything open source Oracle got their hands on, it's been forked.
OpenZFS remains under the widely feared and loathed CDDL 1.O that Sun (not Oracle) applied when they released the source.
Simpler -> Filesystem atop a filesystem atop a filesystem... looks like stacks of BS to me.
Exactly. It's Red Hat trying to have a volume Managing filesystem, instead of going with BTRFS.
Bad choice, python that is, fucking syntax changes every release, it's interpreted so it's slow, all the things any sane person doesn't want in a file system.
The client is python, the server is written in rust. Which part do you think is responsible for handling the actual file system.
Filesystem written in rust and python... meh. You lost me at Python.
The client is python, the server is written in rust. Which part do you think is responsible for handling the actual file system.
Is Rust, with its undeniable benefits, yet stable enough for critical usages? I'm not trolling; I don't use Rust, but have the impression that it still is developing and changing more quickly than I would like as the basis for a production grade file system. Ditto Python.
And then I read this morning that the Red Hat branch of IBM announced they will stop relea0sing RHEL source, which might include that for the Stratus server and supporting software.
Linux is just a Kernel and yes... it sucks
Btrfs had "advantages" not available on other file systems. Benchmark results are perhaps less important.
Ext4 had many advances injected into the Linux kernel, published almost daily by The Linux Foundation. The other two file systems might be useful to enterprises handling many terabytes of information per second. Is this what you are discussing - Enterprise users only?
Most small business and home users are on Microsoft Windows and-or Apple. Does this type of user interest your video channel?
As a Microsoft Windows user, my days is using Microsoft NTFS, with inbuilt Microsoft file compression. Microsoft claims that there is not any benchmark differences. Compression might allow better benchmark results.
Btrfs shows bad benchmark results, in your results, Phoronix and other results. For the small business and non professional users, are these differences significant?
This new file system is interesting? Why? So many claims and press releases do not state the supposed advantages of the new releases. The outsiders like myself are forced to be fearful of these new products. These "advances" are usually buggy, with many incompatibilities with existing settings.
Some products are only alpha and beta releases, but pretend to be ready for use by the general public. Most outsiders therefore will be careful about "Stratis" or whatever new product is released. Is this TH-cam channel another snake oil seller, and how reliable are its claims?
The unbelievable inconveniences folks are willing to go through for "free" and "open" software. At the same time, they don't mind closed and expensive hardware. The open source movement is anti-intellectual at it's core. If something has no physical manifestation, it has no value.
It's all fun ans games until you depend ln someone else to fix stuff, develop the functionality you need or have to live with whatever rules they're pushing down ln you because you own nothing and can't modify anything.
You talk as if the FOSS movement came from folk who just wanted to be cool and whatnot, and not because of necessity.
@@andrebrait actually your statement suggest 2 subjects that are not opposed to each other. The bad people who foisted proprietary software (Billy G.?) were all doing it to STOP innovation being freely distributed.
@@andrebrait What unique value or necessity does FOSS afford us that we didn't have before? The best word processors, spreadsheets, CAD programs and operating systems are still closed and commercial. It has neither increased innovation nor lived up to its spurious claims of protecting user freedoms. In fact, FOSS has lead to an even more onerous trend of software as a service and cloud computing where the user's own data and accessibility are locked up behind a paywall.
Ironically, it is the large software companies, such as Microsoft, Google and Oracle, who have benefited the most from FOSS. Their smaller less capitalized competitors are forced to not only compete against them, but also against inferior FOSS products. There was once a thriving market for shareware and low cost software products developed and supported by individuals and small companies, but they're no longer viable business models.
It becomes no surprise why large companies enthusiastically support FOSS. In addition to solidifying their monopoly, they also get to capitalize on the free labor of eager and passionate FOSS developers under the auspices of FOSS ideals. None of these large software corporations would dare release the source to any of their flagship products, such as Google search or Windows.
FOSS has also decimated the opportunities of the once growing middle class of professional software developers. Despite the astounding growth of the computer industry over past decades, the software developer job market hasn't grown at anywhere near the same pace. FOSS enables the technical infrastructure to outsource software development to lower cost global labor markets. Moreover, FOSS lacks the financial barriers such as license, training and support fees of traditional closed software, eliminating a significant advantage of domestic software developers in a global marketplace.
In summary, FOSS has been an absolute disaster to software quality, the developer economy, and to the very freedoms it purports to champion and protect.
Funny, your use of "anti-intellectual" seems so narrow. Extremely creative people working in Linux to contribute to so many others, seems like a very good thing.
everyone replying here fell hard for the bait