Comments on: Around the Block: High-Performance *nix Filesystems http://heliologue.com/2006/12/18/around-the-block-high-performance-nix-filesystems/ Let joy be unconfined. Let there be dancing in the streets, drinking in the saloons, and necking in the parlor. Wed, 23 May 2012 05:37:09 +0000 hourly 1 http://wordpress.org/?v=3.3.2 By: Ben http://heliologue.com/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36390 Ben Tue, 19 Dec 2006 04:43:40 +0000 http://heliologue.com/blog/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36390 Ah, I should have searched first: some <a href="http://linux.inet.hr/first_benchmarks_of_the_ext4_file_system.html" rel="nofollow">benchmarks</a> from 2 months ago have ext4 generally outperforming ext3 by a fair margin, and even surpassing Reiser4 in some instances. Ah, I should have searched first: some benchmarks from 2 months ago have ext4 generally outperforming ext3 by a fair margin, and even surpassing Reiser4 in some instances.

]]> By: Ben http://heliologue.com/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36381 Ben Tue, 19 Dec 2006 03:44:04 +0000 http://heliologue.com/blog/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36381 Chris, I've never been impressed with anything Reiser has done. Perhaps I'd be singing a different tune in '01, when ReiserFS was the only journaling FS around, but by the time I became interested in Linux, v3 was antiquated and v4 was still a gleam in Hans' eye. Even now, though Reiser4 purports to be the end-all, be-all of filesystems (didn't Hans even say it would take on a lot of functions previously relegated to databases? I think he did...), it has yet to be accepted into the kernel, it's got a pluggable nature which could be the best <em>or</em> the worst thing ever, and its performance isn't <em>that</em> far and away better than what else is out there. Besides, if ZFS gets GPLed like Java is, I think there'll be a new champion of robust filesystems. By the by: funnily enough, the physical block journaling that ext3 does when it's set to <code>data=journal</code> at mount or in the <code>fstab</code> actually makes it <em>faster</em> in situations where mixed reads/writes are constant. So ext3 isn't really such a bad choice for a server after all. What's more, I'm interested to see how ext4 muddies up the waters: I know extents that improve performance, but at the expense of backward compatibility. Ah well, Ts'o hasn't steered me wrong yet. Chris, I’ve never been impressed with anything Reiser has done. Perhaps I’d be singing a different tune in ’01, when ReiserFS was the only journaling FS around, but by the time I became interested in Linux, v3 was antiquated and v4 was still a gleam in Hans’ eye.

Even now, though Reiser4 purports to be the end-all, be-all of filesystems (didn’t Hans even say it would take on a lot of functions previously relegated to databases? I think he did…), it has yet to be accepted into the kernel, it’s got a pluggable nature which could be the best or the worst thing ever, and its performance isn’t that far and away better than what else is out there.

Besides, if ZFS gets GPLed like Java is, I think there’ll be a new champion of robust filesystems.

By the by: funnily enough, the physical block journaling that ext3 does when it’s set to data=journal at mount or in the fstab actually makes it faster in situations where mixed reads/writes are constant. So ext3 isn’t really such a bad choice for a server after all.

What’s more, I’m interested to see how ext4 muddies up the waters: I know extents that improve performance, but at the expense of backward compatibility. Ah well, Ts’o hasn’t steered me wrong yet.

]]> By: Ralph Dagza http://heliologue.com/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36366 Ralph Dagza Tue, 19 Dec 2006 01:00:53 +0000 http://heliologue.com/blog/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36366 Congratulations i saw your site listed on 9rules! Congratulations i saw your site listed on 9rules!

]]> By: Christopher Egner http://heliologue.com/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36337 Christopher Egner Mon, 18 Dec 2006 21:37:03 +0000 http://heliologue.com/blog/2006/12/18/around-the-block-high-performance-nix-filesystems/#comment-36337 Hmm, interesting article. The tests of Resier4 were less than impressive. I've used it in the past, wasn't worth it for me though. Truth be told, for users its probably generally best to stick with ext3. However servers are another story. It'd be interesting to see what hosting services use. Hmm, interesting article. The tests of Resier4 were less than impressive. I’ve used it in the past, wasn’t worth it for me though.

Truth be told, for users its probably generally best to stick with ext3. However servers are another story. It’d be interesting to see what hosting services use.

]]>