A Look At The HAMMER2 File-System Performance With DragonFlyBSD 5.2

[phoronix]

Phoronix: A Look At The HAMMER2 File-System Performance With DragonFlyBSD 5.2

With this week's release of DragonFlyBSD 5.2 this popular BSD operating system is promoting its own HAMMER2 file-system as stable. As a result, here are a few fresh benchmarks of HAMMER vs. HAMMER2 on DragonFlyBSD 5.2 while more tests are forthcoming.

http://www.phoronix.com/vr.php?view=26195

[AndyChow]

HAMMER2 is really good, so far. The biggest problem for me has been the lack of documentation. Also, the whole social justice bs of FreeBSD has caused a lot of really smart maintainers to flee, and since a lot of packages are no longer maintained, they also aren't being flowed from ports into dports.

Nevertheless, DragonFly is the non-linux alternative that actually works. FreeBSD is busy cannibalizing itself, OpenBSD and NetBSD aren't multi-processor efficient.

DragonFly is the BSD that is going to make it. It won the BSD wars. It's just that others haven't figured it out yet.

[jacob]

HAMMER2 is really good, so far. The biggest problem for me has been the lack of documentation. Also, the whole social justice bs of FreeBSD has caused a lot of really smart maintainers to flee, and since a lot of packages are no longer maintained, they also aren't being flowed from ports into dports.

Nevertheless, DragonFly is the non-linux alternative that actually works. FreeBSD is busy cannibalizing itself, OpenBSD and NetBSD aren't multi-processor efficient.

DragonFly is the BSD that is going to make it. It won the BSD wars. It's just that others haven't figured it out yet.

OpenBSD doesn't need to have stellar multiprocessor performance. It's goal is to be an ideal OS for network devices and it's extremely good at that.

I agree that the FreeBSD community has long been full of sh*t. It would certainly be interesting to see Dragonfly make some inroads in the real world and bring some competition to Linux distros. Judging by benchmarks it still has a long way to go before it can rival Linux in SMP performance but the competition between those two radically different approaches will be fascinating.

[Steffo]

HAMMER2 is really good, so far. The biggest problem for me has been the lack of documentation. Also, the whole social justice bs of FreeBSD has caused a lot of really smart maintainers to flee, and since a lot of packages are no longer maintained, they also aren't being flowed from ports into dports.

Nevertheless, DragonFly is the non-linux alternative that actually works. FreeBSD is busy cannibalizing itself, OpenBSD and NetBSD aren't multi-processor efficient.

DragonFly is the BSD that is going to make it. It won the BSD wars. It's just that others haven't figured it out yet.

From the point of industry, FreeBSD and OpenBSD are the BSD derivates used most. In contrast, I've never heard, that a company is using DragonFlyBSD.
DragonFlyBSD might have an interesting file system, but FreeBSD has ZFS which is very mature and production ready.

[waxhead]

Trying out DragonFly right now , who knows! Maybe it is worth using

[hyper]

Back in 2011, Matthew Dillon explained to you the problems with the blogbench benchmark. 7 years later, you don't appear to have paid any attention to what he said, and you really should have.

[carewolf]

HAMMER2 is really good, so far. The biggest problem for me has been the lack of documentation. Also, the whole social justice bs of FreeBSD has caused a lot of really smart maintainers to flee, and since a lot of packages are no longer maintained, they also aren't being flowed from ports into dports.

Nevertheless, DragonFly is the non-linux alternative that actually works. FreeBSD is busy cannibalizing itself, OpenBSD and NetBSD aren't multi-processor efficient.

DragonFly is the BSD that is going to make it. It won the BSD wars. It's just that others haven't figured it out yet.

Isn't DragonFly one of those FreeBSD factions that you are saying are "cannibalizing themselves"?

[AndyChow]

Isn't DragonFly one of those FreeBSD factions that you are saying are "cannibalizing themselves"?

DragonFly isn't FreeBSD. By "cannibalizing themselves", I was talking about the whole code of conduct where they lost some of their most dedicated and talented people, by basically imposing a solution to a problem that didn't exist. It's so bad they no longer have a spice-gtk or xf86-video-qlx maintainer.

The current adoption rate of DragonFly is irrelevant. They have tackled and understand fundamental scaling issues.

Look at how badly FreeBSD behaves with any hardening, or now kpti. ZFS is ok, but slow and hungry. BTRFS is always buggy. HAMMER allows flexible topology and is rock-solid. With the meltdown-spectre issue, the DragonFlyBSD fix was fast and almost trivial, because the basic architecture is actually well thought-out. It's a mainframe class OS.

FreeBSD, now with clang, is just a slow hobby OS. OpenBSD is great, but not future-proof. NetBSD is embedded class.

Obviously, Linux trumps them all. But BSD wise, DragonFly is going to be the supreme leader.

[carewolf]

DragonFly isn't FreeBSD. By "cannibalizing themselves", I was talking about the whole code of conduct where they lost some of their most dedicated and talented people, by basically imposing a solution to a problem that didn't exist. It's so bad they no longer have a spice-gtk or xf86-video-qlx maintainer.

The current adoption rate of DragonFly is irrelevant. They have tackled and understand fundamental scaling issues.

Look at how badly FreeBSD behaves with any hardening, or now kpti. ZFS is ok, but slow and hungry. BTRFS is always buggy. HAMMER allows flexible topology and is rock-solid. With the meltdown-spectre issue, the DragonFlyBSD fix was fast and almost trivial, because the basic architecture is actually well thought-out. It's a mainframe class OS.

FreeBSD, now with clang, is just a slow hobby OS. OpenBSD is great, but not future-proof. NetBSD is embedded class.

Obviously, Linux trumps them all. But BSD wise, DragonFly is going to be the supreme leader.

Okay. I just thought DragonFly BSD was a fork of FreeBSD, which it might once have been, but it sounds like it is a permament fork like the other BSDs the way you describe it.

[phoenix_rizzen]

Okay. I just thought DragonFly BSD was a fork of FreeBSD, which it might once have been, but it sounds like it is a permament fork like the other BSDs the way you describe it.

DFlyBSD started as a fork of FreeBSD 4.x to try a different path to better SMP scaling compared to FreeBSD 5+, with a side goal of single-system image clustering. It has since developed into it's own OS that's quite different from FreeBSD 11 of today. The SSI clustering hasn't emerged yet, but HAMMER is being geared toward that goal, and is getting closer with HAMMER2.

DFlyBSD is now it's own, full-fledged BSD OS, separate from (but derived from) FreeBSD. Similar to how OpenBSD forked from NetBSD to become it's own full-fledged OS.