Alternative file systems? Was Re: Better File systems?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, 2002-04-18 at 14:50, Richard Troy wrote:
> 
> cry for "performance." My rethort was, "Yeah, some performance while
> you're rebuilding your system from tape! Hah!"

:-)

> So, I think it'd be really neat if someone would put together a small
> table outlining what's available for our beloved RedHat Linux
> systems, telling us whether or not they're included with Red Hat, or
> whether we have to hunt down the RPMS, etc., and perhaps with a brief
> statement of the technology and good applications. What I'm thinking of

I think i already did that. But anyway...

These are the journalised FSs that i know:

1. Ext3
Its main advantage is the easy upgrade from Ext2. Now, if you're like me
and never upgrade but, instead, reinstall from scratch, this advantage
looses much appeal.
Nevertheless, it's good because "it's like ext2", people are familiar
with it, there's a good chance that applications will work well with it
(because they are already working well with Ext2)...
When you're interested in the safety of the data, not only metadata, use
it with data=journal. (but i can only think of databases in this case,
and databases work actually better on raw disks)
Performance-wise, there's a strange situation with some mail servers,
when the performance on Ext3 with data=journal is actually very high! It
has to do something with the order of the sync writes.
It's available in the "official" RH distribution.

2. ReiserFS
It is supposed to have a very high performance with lots of small files
being created and deleted all the time, and that makes it ideal for
proxy servers (Squid).
It's fairly recent, and i had some stability problems with it one year
ago, especially under heavy load and lots of disk I/O. But things might
have been improved meanwhile.
I guess there are more interesting things coming from ReiserFS in the
future. It looks like its going to be a very featureful and clever FS
(it tries to blur the difference between the FS and a database, and
other nifty tricks).
The "official" RH kernels do have ReiserFS support, but suboptimal:
debug is turned on. I guess you can find modified
kits/installers/kernels on the Net if you search them. I think some
brasilian guys used to offer a ReiserFS-enabled RH...

3. XFS
This is interesting, because the XFS code base is actually very mature
and stable. That's why, i guess, the Linux port became stable so
quickly: because only the Linux "hooks" had to be made stable, while the
core was already mature.
>From my experience (and from what i've been told by other people trying
to solve similar problems), this is the FS that offers the best
stability under extreme loads (in terms of CPU and I/O), especially for
fileservers, or for (concurrent) access to very large files, or when you
need "guaranteed" disk I/O bandwidth. Actually, it was designed to be
like that: SGI needed those features for its 512 CPU video-processing
machines. ;-)
You can do Posix ACLs with it, and Samba takes advantage of that,
translating them into Windows "ACLs". Quota also works. It also has
extended attributes that enable you to store customised extra file
information at the FS level (think of BeOS filesystem).
Previous releases (up to 1.0.2) were slow when it comes to deleting lots
of files, because all deletes were synchronous; this thing made it
somewhat unappropriate for proxy servers and mail relays. The latest
stable release (1.1) fixes that (although some people still believe sync
deletes are "the right way" ;-)).
You can get an XFS-enabled RH installer from here:
http://oss.sgi.com/projects/xfs/

-- 
Florin Andrei

There's nothing to be ashamed of in coming up with the obvious,
especially when nobody else is coming up with it.





[Index of Archives]     [Kernel Newbies]     [Red Hat General]     [Fedora]     [Red Hat Install]     [Linux Kernel Development]     [Yosemite News]

  Powered by Linux