You are not logged in.
Pages: 1
i just read this
http://www.linux-mag.com/cache/7345/1.html
and it looks like the performance benefits of NILFS are extremely promising. not to mention the snapshot ability.
NILFS support should be built into the 2.6.30 kernel now which is what the most recent arch release uses (2009.8). I plan to install Arch on my laptop (lenovo x200) it didnt work with older versions of arch but hopefully it will work now.
i dont think the installer supports NILFS yet (the install guide doesnt have it in there) is there a way to try it out? preferably on the root partition. a lot of people use reiserfs for /var i think because it is fast for small files but NILFS seems to be way way faster so it may be even better for /var at least. using it for /home would be nice too for the snapshot ability.
the article does mention that there are issues with SSDs because of a large number of writes; i have an SSD. this report was written in June so i dont know if that has been resolved yet. does anyone know?
this is an extremely interesting technology though. im very excited for this to hit the forefront and make linux the king of filesystems in the way that solaris has been with ZFS.
thanks.
Offline
Install in the usual way, then change to nilfs when the system is up and running.
Offline
1) Install Arch with one of the supported filesystems (If you don't have Arch set up already) .
2) Create a sample nilfs2 partition and mount it.
3) Read the 2 warning lines and copy/paste them here .
4) Use the sample partition for testing purposes or delete it .
5) Move on .
Edit : you need nilfs-utils for mounting of course . It's preferable that you build the latest upstream release (the package in community is outdated) .
Last edited by Nezmer (2009-10-08 21:58:07)
English is not my native language .
Offline
Edit : you need nilfs-utils for mounting of course . It's preferable that you build the latest upstream release (the package in community is outdated) .
is not the case anymore. i've updated to .14
Give what you have. To someone, it may be better than you dare to think.
Offline
Nezmer wrote:Edit : you need nilfs-utils for mounting of course . It's preferable that you build the latest upstream release (the package in community is outdated) .
is not the case anymore. i've updated to .14
Ouch .
In my defense , It didn't reach the mirrors yet .
English is not my native language .
Offline
Out of interest... what 2 warning lines?
Allan-Volunteer on the (topic being discussed) mailn lists. You never get the people who matters attention on the forums.
jasonwryan-Installing Arch is a measure of your literacy. Maintaining Arch is a measure of your diligence. Contributing to Arch is a measure of your competence.
Griemak-Bleeding edge, not bleeding flat. Edge denotes falls will occur from time to time. Bring your own parachute.
Offline
# mount.nilfs2 -v -o users /dev/sdb1 /mnt/4GB
mount.nilfs2: WARNING! - The NILFS on-disk format may change at any time.
mount.nilfs2: WARNING! - Do not place critical data on a NILFS filesystem.
mount.nilfs2: started nilfs_cleanerd
/dev/sdb1 on /mnt/4GB type nilfs2 (rw,noexec,nosuid,nodev,gcpid=5133)
Those
edit: They also come without the -v
Last edited by Army (2009-10-09 07:16:39)
Offline
By the way, the last time I heard about nilfs, it had pretty awful performance. Although I'm pretty sure they're working on it, I don't know how far did they get.
Offline
# mount.nilfs2 -v -o users /dev/sdb1 /mnt/4GB mount.nilfs2: WARNING! - The NILFS on-disk format may change at any time. mount.nilfs2: WARNING! - Do not place critical data on a NILFS filesystem. mount.nilfs2: started nilfs_cleanerd /dev/sdb1 on /mnt/4GB type nilfs2 (rw,noexec,nosuid,nodev,gcpid=5133)
Those
edit: They also come without the -v
Ouch! Scary stuff there.
Allan-Volunteer on the (topic being discussed) mailn lists. You never get the people who matters attention on the forums.
jasonwryan-Installing Arch is a measure of your literacy. Maintaining Arch is a measure of your diligence. Contributing to Arch is a measure of your competence.
Griemak-Bleeding edge, not bleeding flat. Edge denotes falls will occur from time to time. Bring your own parachute.
Offline
I'll play around with it on my usb stick, for using it on / I'm not insane enough ;-)
Offline
I gave NILFS a try (NILFS2 actually) installing it on the root partition and it looks good )
Here my experiences: http://www.blah-blah.ch/Mra/Nilfs2performance
Offline
I gave NILFS a try (NILFS2 actually) installing it on the root partition and it looks good )
Here my experiences: http://www.blah-blah.ch/Mra/Nilfs2performance
maybe comparisons with btrfs and ext4 are more appropriate?
ᶘ ᵒᴥᵒᶅ
Offline
SSD optimized filesystem (~NILFS) is on the roadmap.
http://wiki.archlinux.org/index.php/Dev … ng_roadmap
we only need someone who wants to do the work
< Daenyth> and he works prolifically
4 8 15 16 23 42
Offline
unodue wrote:I gave NILFS a try (NILFS2 actually) installing it on the root partition and it looks good )
Here my experiences: http://www.blah-blah.ch/Mra/Nilfs2performancemaybe comparisons with btrfs and ext4 are more appropriate?
Maybe a little in the case of btrfs (if the filesystem is mounted with the ssd options), otherwise I don't think so - they're both journaled filesystems in the end meaning that for each file they'll write they'll have to modify the journal (therefore falling into the slow flash-modify cycle) => slowdown on flash.
But ok, for the sake of confirming this I'll do such a comparison once I'm back from my holidays )
Offline
Maybe a little in the case of btrfs (if the filesystem is mounted with the ssd options), otherwise I don't think so - they're both journaled filesystems in the end meaning that for each file they'll write they'll have to modify the journal (therefore falling into the slow flash-modify cycle) => slowdown on flash.
But ok, for the sake of confirming this I'll do such a comparison once I'm back from my holidays )
i'd be very interested in the amount of background i/o with nilfs2 versus btrfs. because of nilfs' log-structured nature, there are reports that sometimes a multitude of data is written for seamingly simple operations. this would ofcourse have an adverse effect on ssd lifespan, so it would be interesting to investigate this further.
see also this (older) post: http://www.ocztechnologyforum.com/forum … post365779
ᶘ ᵒᴥᵒᶅ
Offline
I have 2 SSD drives. I installed opensuse with btrfs on one yesterday and i have arch linux with ext4 on the other one. the opensuse one is slc and the arch one is mlc.
I would be happy to help out in any way i can; i want to give something back to the community.
I am not sure though how to really compare btrfs and nilfs (and any others) in terms of wear-leveling. the nature of a log structured file system leads to a lot of writes over the drive but btrfs from my understanding writes very heavily to particular locations which will also have adverse effects.
let me know if there are anything anyone wants me to try; im really interested in fs theory.
Offline
But ok, for the sake of confirming this I'll do such a comparison once I'm back from my holidays )
So, I ran the same tests as well against btrfs and ext4 using another USB-stick but still from the same manufacturer and exactly the same model (Trascend JF V30 16GB) - here the results: http://www.blah-blah.ch/Mra/Nilfs2performance
Hope it's useful!
I am still sticking to nilfs2 for my root partition
Bye!
Offline
Pages: 1