ralf@uni-koblenz.de wrote:
>
> Hans,
>
> I don't know if you're following linux-kernel@vger.rutgers.edu. There
> has been quite some talk about possible extensions to ext2 during the
> last time on this list. Given that you are working on a new filesystem
> that is supposed to be superior compared to ext2 in most, if not all
> disciplines I'm asking myself the question if there is still sense in
> doing serious development work on ext2. Therefore my question, how long
> do you think until Reiserfs has reached production quality? I'm of course
> aware that this can only be an estimate but based on this estimate I think
> developers could decide whether the investment of time into ext2 makes
> sense or not.
>
> Another thing which I think would be interesting is comparing Reiserfs to
> SGI's XFS. I'm using both ext2fs (Linux/MIPS) and XFS on an Indy and have
> to admit that XFS has a lot of killer qualities. The general implementation
> of Linux for Indys is not yet advanced enough to make a comparison of XFS
> vs. ext2fs on the same machine a fair one.
>
> Ralf
I just subscribed to linux-kernel, I should have subscribed years ago
rather than looking in the archives every couple of months as I have
done.
We hope to have our second beta ready for beta-release by the end of the
month, and a fully released file system proven in its stability in
June. Our second beta is able to copy the linux kernel, but it is not
ready for release yet. Our previous beta reached the point where it
required some effort to crash, but Linus made some remarks about our
interface to VFS and how we should toss leashes, accurate remarks that
required our scrapping a remarkable amount of code to fix (the code is
much cleaner now as a result), and we improved it accordingly. The new
dcache code looks like hot stuff to us, I am hoping it will improve our
benchmarks further.
If the ext2 guys were to help us with such things as tweaking the
NFS-reiserfs interaction, improving reiserfs, adding some of their
optimizations such as what they have done in the area of read-ahead,
etc., that would help much.
I honestly have no idea how we will compare with XFS. I have a lot of
respect for their design. I expect that they would do better at large
I/O simply because they have tweaked their buffering system to work with
XFS. This is not any inherent advantage in their approach, just a
description of what we have not done yet. I have some ideas for how to
do it for Reiserfs that I would like to explore someday. There are more
things we could do to optimize million file directories that I wonder if
they have already done. We should use full filenames in keys....
Thanks Ralf,
Hans
--------------8A6E76CC607F87DC75D45E59
Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Hans Reiser
Content-Disposition: attachment; filename="vcard.vcf"
begin: vcard
fn: Hans Reiser
n: Reiser;Hans
org: The Naming System Venture
adr: 6979 Exeter Dr.;;;Oakland;CA;94611;USA
email;internet: reiser@ricochet.net
title: Owner
tel;work: 510-482-2483
tel;home: 510-482-5071
note: Phone: +1 (510) 459-4681 (cell phone)
x-mozilla-cpt: ;0
x-mozilla-html: TRUE
version: 2.1
end: vcard
--------------8A6E76CC607F87DC75D45E59--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu