On Sun, May 06, 2001 at 06:24:12PM +0800, Federico Sevilla III wrote:
> Does anyone have information on performance? Space utilization?
reiserfs with tails enabled (default) should be more space-efficient.
When it comes to performance I belive it depends on what kind of load
you're using it for. I believe reiserfs will be faster for lots of small
files, but XFS will be faster for streaming big files.
I'll post real numbers once we've run som real tests though.
> Reliability when recovering from unclean shutdowns?
>
> In the area of reliability when recovering from unclean shutdowns I'm
> particularly interested at how XFS deals with its cache/buffer. For
> example in mail queues, with ext2 directories can be flagged using chattr
> as +S so that all work done with these directories are synchronous. A
> little slower, but this is much more stable especially for these queues.
> With ReiserFS AFAIK mail queues currently run on a risk because perfectly
> timed power downs could cause mail to "disappear". How would situations
> like this be handled by XFS?
I think this is a misunderstanding.
All three filesystems will guarantee you that the filesystem is intact
after a crash, but data isn't journaled (except with ext3 and
data-journaling enabled), so it doesn't ensure that the _content_ of the
files are up-to-date. Your application should use fsync to ensure this.
(chattr +S on ext2 is just a different way of using fsync)
A MTA should write the mail to disk (and use fsync) _before_ it tells
the sender that the mail has been accepted. This way there is no risk of
loosing mail. I believe qmail doesn't use fsync, but relies on the
filesystem beeing mounted with sync-option (or chattr +S on that
directory). Because doing all IO syncroniously would kill performance,
the reiserfs-team fixed qmail instead of suggesting that people mount
with sync option (or implement chattr +S).
--
Ragnar Kjørstad
Big Storage
|