On Mon, 5 Jun 2000, Hans Reiser wrote:
> Chris Mason wrote:
> >
> > [ cannot compare the journals by benchmarking ext3 and reiserfs ]
>
> I disagree, performance is the primary, though not the only, measure of
> journaling
> code.
>
Ok, I'll try again. We have two journal layers. We want to compare their
speed, and their usefulness in the kernel as a whole. To compare the
performance, we have at least two choices:
1) Port the journal layer from one FS into the other, or port both journal
layers into a new, benchmarking fs.
2) Compare the overhead of each journal relative to the native FS.
You cannot compare ext3 with reiserfs use that to benchmark the journal
layers. There is a slightly different FS beneath the two journal systems,
and I suspect it might affect the results ;-)
As for their usefulness to the kernel as a whole, one tries to provide a
generic journal facility, and the other doesn't. That doesn't make the
reiserfs log bad, but I do understand the goodness factor of the
JFS layer.
> Alan wants us to wait for a committee that will sit around and build a giraffe.
Alan wants us to be a part of the group that finds the right way to
put all the new filesystems in the kernel. He is thinking about what it
will take to maintain things in the long run, and in the end it will make
us a better filesystem.
We don't have to agree with him, but we do have to deal with criticism
without going off the deep end.
Right now, the reiserfs code has a very small footprint in the main kernel
source, and we've killed most of the duplicated code in buffer2.c (more
can go out, actually). I don't think we will get in the way of future FS
integration.
> I want him to stop using FUD about format changes,
The fears about format changes are our fault. Some of our own users don't
understand what works where, because we haven't been consistent in our
mail or our documentation. Just count the number of clarification email
messages we sent out yesterday. All we can do is try to be more clear in
the future...
-chris
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:25 EST