Reiser is better....why?

[dkmweeks]

I adopted the Reiser filesystem when I adopted Gentoo. It is supposed to be better, but hdparm not with standing, I've got some serious performance problems with this file system.

Or so I assume.

What I know is that some (too much) of my disk access has really poor performance.

What I don't know is why Reiser is so much better than ext2/3, regarding throughput.

David

[Rene-dev]

is dma enabled?
here are some benchmarks
http://www.namesys.com/benchmarks.html
http://vizzzion.org/?id=reiser4
or what about this great site?
http://www.google.com/search?hs=2CW&hl=en&lr=&client=firefox-a&rls=org.mozilla%3Aen-US%3Aofficial&q=reiserfs+ext3+benchmark&btnG=Search


I use ext3

rene

[Archangel1]

hdparm will narrow down some issues, but it's test is so contrived it's not necessarily indicative of real-world stats - and it won't test the filesystem.

The reason Reiser is better than ext3 is fairly involved. For a start it's a lot newer; ext3 = ext2 + journalling, and ext2 is quite old now. Hence Reiserfs uses some more advanced algorithms to get better performance; this is at the expense of an increase in CPU use, so if you're installing on a P100, ext2/3 would likely be a better choice.

As I understand it, reiserfs' main advantage is in small files; this is because it's capable of storing small quantities data in the tree itself, rather than just a pointer to it.
Hence when you want to read bunches of small files, it can simply read through the tree, rather than reading part of the tree, then the file it points to, then the next part of the tree, etc. Given hdd seek times this results in a significant improvement.

Reiser4 is meant to be sort of like reiserfs but more so; I don't really understand much of what's going on there. It's a controversial topic anyway so I'll leave it alone
_________________
What are you, stupid?

[dkmweeks]

Thanks for the replies.

I've tweaked my setting with hdparam, including dma.

I suspect part of my problem is fragmentation, but I don't readily know how, or even if I should, check that. My "lightweight" understanding is that ext2/3 and Reiser (I'm assuming) are self correcting regarding fragmentation.

This drive (laptop) gets full, and then I move stuff onto a server. I also use ccache, and at one point had to rm the cache to make space. So I know there's been lots of opportunity for fragmentation.

Are there any cron worthy utilities you'd recommend?

David Weeks

OH!

One other thing comes to mind: single process streaming performance is one thing, but I run lots of stuff on this laptop. What do you know about the effects of divided drive access requirements?

[ph03n1x]

I'm also using reiser-fs on a laptop and I'm happy here

Performance is good so I guess it's not a reiser issue. Pros for Reiser are that it can handle small files as good as big ones it needs afaik no defragmentation because that has not such an impact as for windows fs. Ask google it knows a lot

Of course if your drive is always nearly full that will slow down the hd extremly. Also think of the fact that most Laptop Hds are only 5400rpm and around 12ms or so...

[Archangel1]

[yoshi314]

my personal experience with ext3 and reiserfs:

ext3 seemed somewhat slow and disk worked almost as loud as with windows (lots of weird "tick"-like sounds that disks do when seeking data)
reiserfs is faster by a certain degree and my drive works a lot more quiet somehow (i had real difficulty to hear mentioned "ticks")

also journal replay was faster, and general portage operations were done better. (like unpacking kernel source)

i switched from reiser to ext3 once (buggy baselayout seemed to wreak havoc on my reiser partition) packing / into tar and then untarring it back (after mkfs) . later on i switched back to reiser in the same manner. so it seems like a fair comparison.
_________________
~amd64
shrink your /usr/portage with squashfs+aufs

[dkmweeks]

Thanks everyone.

I think my situation comes down to a heavily loaded laptop. I've not had a problem with Reiser (edit: except for capacity related issues, and some really slow mv and cp.), and YES, the journaling is much better than ext2 + journaling.

I see now that there is a 120 gig, 2.5" laptop drive available..... hmm...

Dave

[Gentree]

From the vizzzion.org link above:

[GaMMa]

I've been using Reiser4 for probably a year now and I think it's a great filesystem. The drive is quieter compared to ext3, kernel sources decompress faster and it just seems faster overall. I've got a 5400 laptop drive and ext2/3 used to make it crawl, but reiser4 seems a lot better for the drive.
_________________
Ubuntu Linux Dapper Drake running Gnome-2.14.1
[Website | Screenshot | Portage Guide]

[Muhiz]

I used Reiser4 over 9 months but recently I changed back to other file systems like JFS, XFS and ReiserFS3. I had two major corruptions while using R4 and it's bit problematic to get a kernel, which works with R4 and ATI's binary drivers. Those corruptions were linked with some library incompatability with LVM2 ( missing some lib-files, while loading LVM2 ). Although I never lost any data but I got tired for various problems. Reiser4 is unbelievable FS: it's really fast, it uses 25% less HD space than R3 and almost half as much as competitive file systems ( I may remember those numbers wrong but it's true ).
_________________
-- Muhiz --

[nxsty]

[asiobob]

R4 is nice and all. Except their plugin feature is an attempt to re-invent the wheel (READ the kernel Virtual File System sub section).
Until we get our ego's in the right place I can't see R4 getting into the kernel. R4 basically nearly goes around the in-kernel VFS section.

IT doesn't even have extended attributes in the sense of other file systems (it does but not compatible). It renders apps like beagle uber slow since it has to reply on sqlite.

If the guys and gals at namesystem sorted this issue out that would a tad awsome -- just like the rest of R4 which is indeed great.

EXT3 has a zillion people look over the code and use it. It's tried and tested of course.