nilfs2 per directory quota

nilfs2 per directory quota

[Niccolò Belli]

Hi,
I read nilfs2 does support per directory quota, but I didn't find how to
achieve it: the few peoples who need directory quota do use xfs.
Can you please give me some more info? Is nilfs2 still too much immature
for a production environment?

Cheers,
Darkbasic
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: nilfs2 per directory quota

[dexen deVries]

Hi Niccolò,


On Sunday 26 June 2011 11:48:32 you wrote:
> I read nilfs2 does support per directory quota, but I didn't find how to
> achieve it: the few peoples who need directory quota do use xfs.

casually looking through NILFS2 sources, I don't see quota support, but I may 
be wrong there. Why isn't XFS a viable option for you, if I may?


> Can you please give me some more info? Is nilfs2 still too much immature
> for a production environment?

I'm using it under light-to medium load for a few months now, both on desktops 
and test server. In v2.6.39 it seems rock solid -- but that's just hear-say 
evidence.

On caveat applies: under certain workload, current implementation suffer from 
high file fragmentation. That workload for me is Bitcoin, which modifies a lot 
of data in Berkley DB with pagesize 4kB. Also, BDB in Bitcoin issues fsync() 
or similar after every disk write, which creates lots of checkpoints, eating 
up diskspace with metadata.

But that's an extreme case; for general use NILFS2 is a-OK.


Regards,
-- 
dexen deVries

> (...) I never use more than 800Mb of RAM. I am running Linux,
> a browser and a terminal.
rjbond3rd in http://news.ycombinator.com/item?id=2692529
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: nilfs2 per directory quota

[Niccolò Belli]

Hi,

Il 26/06/2011 13:46, dexen deVries ha scritto:
> casually looking through NILFS2 sources, I don't see quota support, but I may 
> be wrong there.

I read about per directory quotas in slashdot:
http://linux.slashdot.org/comments.pl?sid=1287319&cid=28536529


> Why isn't XFS a viable option for you, if I may?

XFS *is* a viable option, but it was a chance to test a new fs ;)


Cheers,
Darkbasic
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: nilfs2 per directory quota

[dexen deVries]

Hi again,


On Sunday 26 June 2011 13:58:45 you wrote:
> Il 26/06/2011 13:46, dexen deVries ha scritto:
> > casually looking through NILFS2 sources, I don't see quota support, but I
> > may be wrong there.
> 
> I read about per directory quotas in slashdot:
> http://linux.slashdot.org/comments.pl?sid=1287319&cid=28536529

I'm affraid the meaning of the post is ``I wish NILFS provided those:'' (the 
``but how about having some real features'' bit).


> > Why isn't XFS a viable option for you, if I may?
> 
> XFS *is* a viable option, but it was a chance to test a new fs ;)

I've been using XFS for several years, also in configuration with external 
journal for extra performance. NILFS2 comes close, if not beter, in 
performance in some cases; mostly for writting new data. For example, recently 
checked out all linux kernel sources on a 5.400 RPM laptopd hdd, and it was 
pretty darn fast. On the other hand, simply traversing the filesystem when 
metadata is not cached can be slow.

Current implementation of NILFS2 has certain on-disk inefficiences in regard 
to metadata, for example it doesn't use extents. The folks at NTT are working 
on improving stuff.




You may also want to check out
http://en.wikipedia.org/wiki/Fossil_(file_system)
http://en.wikipedia.org/wiki/Venti
which are available on any POSIX  environment via 
http://en.wikipedia.org/wiki/Plan_9_port

I don't think Fossil provides any quota support, but when coupled with Venti, 
it pretty much removes the need for it :P


Regards,
-- 
dexen deVries

> (...) I never use more than 800Mb of RAM. I am running Linux,
> a browser and a terminal.
rjbond3rd in http://news.ycombinator.com/item?id=2692529
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: nilfs2 per directory quota

[Niccolò Belli]

Il 26/06/2011 14:48, dexen deVries ha scritto:
> I'm affraid the meaning of the post is ``I wish NILFS provided those:'' (the 
> ``but how about having some real features'' bit).

You're right, the expectations were so high that I assumed it was an
implemented featured :(

> You may also want to check out
> http://en.wikipedia.org/wiki/Fossil_(file_system)
> http://en.wikipedia.org/wiki/Venti
> which are available on any POSIX  environment via 
> http://en.wikipedia.org/wiki/Plan_9_port
> 
> I don't think Fossil provides any quota support, but when coupled with Venti, 
> it pretty much removes the need for it :P

I also read about fusequota (http://code.google.com/p/fusequota/) but it
seems more like a personal experiment rather then something you can use
in a production environment (and it seems no longer developed too).

Cheers,
Darkbasic
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html