Btrfs To Ship Multiple Performance Improvements In The Next Linux Kernel

[ssokolow]

It can, and does, run Linux binaries. You can even install apt on it and use Ubuntu or Debian repos.

I'm familiar with what Android is capable of. Your argument has the same flaw as claiming that iOS and MacOS are the same thing because they both use the XNU kernel... or claiming that, because WSL exists, the kernel is the only thing keeping Windows from being unarguably considered a Linux distro.

[treba]

What world I am living in you ask?

How sweet. What about

https://w3techs.com/technologies/comparison/os-linux,os-windows

https://www.zdnet.com/article/linux-now-dominates-azure/

Don't know what happens in the Netherlands, but it certainly doesn't reflects global website shares.

[jacob]

People use "Linux" to mean two things: The kernel and, for lack of an equally concise alternative, the platform. Android doesn't meet that second definition because it can't natively run "Linux" binaries and aims to be a very distinct platform of its own.

It can, and does, run Linux binaries. You can even install apt on it and use Ubuntu or Debian repos.

[ssokolow]

What does that even mean? Linux is a kernel, nothing more, nothing less. "Running Linux" means precisely and only that, running an OS whose kernel is Linux. Android users "run Linux" exactly in the same way as Ubuntu, Fedora or Debian users. Most distros have their own modified kernels too, by the way. Especially in the case of Ubuntu these modifications are particularly extensive to the point of the distro being partly incompatible with upstream kernels.

People use "Linux" to mean two things: The kernel and, for lack of an equally concise alternative, the platform. Android doesn't meet that second definition because it can't natively run "Linux" binaries and aims to be a very distinct platform of its own.

[jacob]

If you had bothered to check, you'd notice that IBM started it's support for Linux back in 1999. Oct 1998 FreeBSD 3.0-RELEASE with initial SMP support was released. Which was quite problematic. By the time of 4.0-RELEASE which had fixed these issues (Spring 2000), train had passed.

You seem fixated on SMP but by 1999 Linux's SMP support was no better than FreeBSD's (Big Kernel Lock etc.) But it was far from being Linux's only, or even the most pressing problem. Both Linux's virtual memory manager and TCP/IP implementation suffered from severe performance problems. Multithreading support was barely usable with the LinuxThreads package being a quick&dirty hack. Access control sucked, with no support for ACLs or MAC. The kernel lacked some important system calls, like sendfile() (later replaced with splice() & friends) or epoll(). The only production-grade filesystem was still ext2, with its capacity limitations and the need to run fsck. In fact IBM's first major contributions were not even in the SMP area. They set to implement a new and proper support for multithreading with their Next Generation POSIX Threads (NGPT) project and they merged JFS as a high-end filesystem. Incidentally, neither of those things really went anywhere. Ingo Molnar's New POSIX Threads Library (NPTL) was eventually selected over IBM's NGPT, and JFS, while becoming mainline, never gained much traction on Linux, probably due to the fact that it came with virtually no maintenance or administrative tools at all. The filesystem problem, which was seen as one of the major weaknesses holding Linux back, was only solved by the arrival of XFS contributed by SGI, and, later, ext3/4 as a direct upgrade path for ext2 users. IBM did work on SMP optimisation big time, but it wasn't immediate and it wasn't the number one project. So claiming that Linux left FreeBSD behind purely because FreeBSD's SMP support was mediocre is a very bizarre interpretation.

You understand licenses quite wrong. BSD license does not denounce ownership/authorship, it's just giving user more freedom to do with the software as they please. Author remains author and he/she could sue you, if you removed the relevant authorship-headers from sources. You could fork it and relicense your fork under GPL but the original author has to remain that.

Talking about who's wrong, neither the BSD licence nor the GPL affects users in any way. They only govern redistributors and downstream developers. The GPL says do what you want with it, but it must stay GPL. The BSD licence says do what you want, and if you want to turn it into a proprietary product, that's fine by me. That's the crux of the problem. Releasing something of value under the BSD licence means automatically giving it for free to proprietary developers. And in fact, it happens all the time. IIRC the first TCP/IP implementation in Windows NT was ported over from ***BSD, where MS completely re-appropriated it under its own closed and proprietary licence (yes, the attribution clause, yada yada, but that's of little consequence when the code is closed source, isn't it).

That's a "freedom" lots of open source contributors will happily do without, thank you very much. For some reason the BSD community is strongly attached to that and that's their right. I don't expect them, or you, to actually try to understand the GPL's logic, but no-one can deny that Linux has thrived under the GPL. So I don't see anything to suggest that Linux would have done better under another licence, especially since while there are many open-source OSes, Linux is by far the most successful of them all.

bah.. 75% mobiles run Android, not Linux. Android is just using Linux kernel and highly modified one at that. Following that same chain of logic one could claim that Android is also BSD because equally important OS component (It's C library) has BSD origins.

What does that even mean? Linux is a kernel, nothing more, nothing less. "Running Linux" means precisely and only that, running an OS whose kernel is Linux. Android users "run Linux" exactly in the same way as Ubuntu, Fedora or Debian users. Most distros have their own modified kernels too, by the way. Especially in the case of Ubuntu these modifications are particularly extensive to the point of the distro being partly incompatible with upstream kernels.

What world I am living in you ask?

Thank you for the answer. So now we know, you live in a world where the Netherlands equate the world and server usage equates AD and SMB servers for Windows clients on corporate networks.

Nice try though.

[ssokolow]

I could send that argument back to you: Nothing stops Linux enthusiasts here from generalizing based upon W3Techs "Comparison of the usage of Linux vs Windows for websites" and present it as a solid proof of overwhelming general Linux rule of ALL servers. When in fact it only reflects operating systems of web servers and nothing more, nothing less. And it has been done here in Phoronix countless times. Ignoring various kinds of servers and other devices sitting behind firewalls in Intranet, disconnected from Internet. The link I posted does count these normally "invisible" machines in.

As I see it, in the today's global economy, margins are extremely tight and that creates automatic similarity in choices made by companies across countries with similar free economies. With government structures being the sole exception because there other factors might be preferred - like efficiency being sacrificed to stronger security. Even there, similar choices might be made, compared to government structures of other countries.

One party or group commonly making one mistake has no bearing on methodological flaws or misrepresentation by another party or group. They're wrong too.

[aht0]

Well, for one thing, historically, there has been quite a bit of divergence in software and hardware buying decisions between North America, Europe, and the Asia-Pacific region for a mixture of economic and cultural/historical reasons.
It's unwise to generalize about the world from one country for that reason alone.

I could send that argument back to you: Nothing stops Linux enthusiasts here from generalizing based upon W3Techs "Comparison of the usage of Linux vs Windows for websites" and present it as a solid proof of overwhelming general Linux rule of ALL servers. When in fact it only reflects operating systems of web servers and nothing more, nothing less. And it has been done here in Phoronix countless times. Ignoring various kinds of servers and other devices sitting behind firewalls in Intranet, disconnected from Internet. The link I posted does count these normally "invisible" machines in.

As I see it, in the today's global economy, margins are extremely tight and that creates automatic similarity in choices made by companies across countries with similar free economies. With government structures being the sole exception because there other factors might be preferred - like efficiency being sacrificed to stronger security. Even there, similar choices might be made, compared to government structures of other countries.

[ssokolow]

Any particular reason to believe that MS software in Netherland has gotten some sort of special artificial advantages over open source software? If no, it should reflect your average country with free economy well enough to extrapolate.

Well, for one thing, historically, there has been quite a bit of divergence in software and hardware buying decisions between North America, Europe, and the Asia-Pacific region for a mixture of economic and cultural/historical reasons.

It's unwise to generalize about the world from one country for that reason alone.

[kreijack]

kreijack lets think about this scenario. Raid 6, 5 Disks and one Disk failed. Parity on one disk is invalid. How does btrfs fix this problem now? How does it know which parity is invalid? Does the restore check it that the file was invalid with parity A and test again with parity B?

This is one of my concerns.

You are describing a case where two disks failed. Because the RAID6 is tolerant up to 2 fault disks, you are fine.
In your case you have two bad disk, and the math says that three disks are enough to rebuild the data.

The key is that BTRFS always checks the data against the checksum, so it is capable to verify if the (re-build) data is valid or not. If not it tries another set of disks, until a valid disks combination is found.
To be more explicit: if the parity is used and it is corrupted, the re-build data doesn't match the checksum; then two options are available:
a) if another set disks is available it is tried, and the loop restarted (i.e. the check against the checksum is re-performed)
b) otherwise BTRFS returns error and the data is not returned from BTRFS to userspace.

[aht0]

Note that it says "The Netherlands". You never want to start an argument about the world by generalizing research on a single country.

Any particular reason to believe that MS software in Netherland has gotten some sort of special artificial advantages over open source software? If no, it should reflect your average country with free economy well enough to extrapolate.