i see the following problems in the wc structure:
- locking
- 10 times as many entries in .svn, as resources
managed
running "svn up" basically blocks a state-of-the-art-p4 for two
minutes, if your repository contains 20.000 entries. "svn st" blocks
for 40 secs. trying to have the working copy on a network share is
impossible. if we do a ls -lRa | wc, we get well over 200.000
entries.
imports are not an issue, they happen to rare and don't bother
every-day users.
a possible solution for locking:
- write ONE lock file with ONE line stating
the relative path and the command run, recurse y/n.
the location of this file could be the
top-level .svn, or some tmp dir ...
--> immediate speed gain for cleanup/up/st
should be 50%
--> simple to implement
possible improvements for .svn:
- eliminate the "auth" completely (i think
this is already an issue), no way this
can be in the wc, this is a client config
parameter
- make the structure flat, eliminating
all directories (=50% improvement in wc)
- there seems room for eliminating other files
too
- maybe there is room for eliminating some
not needed information too, which would
be wiser handeled by a "company caching
proxy" as the one started by justin.
--- Brandon Ehle <azverkan@yahoo.com> wrote:
> solo turn wrote:
>
> >i'd like to suggest to add an issue for speeding up the client. it
> >would nicely fit in the 0.19 performance milestone.
> >
> >
> Some of the performance issues won't be visible until the
> scalability
> stuff is taken care of, so it makes sense to seperate the two by a
> couple of milestones.
>
> >are the goals:
> >- 10 times as fast as up to now
> >
> During import & checkins, ra_dav needs about a 10x speed increase
> to be
> as fast as ra_local & ra_svn. Beyond that I think issue 913 is
> going to
> make the biggest jump in speed (at least for high speed connection
> operation) and we should see a 2x to 3x speed increase.
>
> >- linear speed curve with entries dealt with
> >utopic?
> >
> >
> Subversion's pretty good at that right now, I don't see any
> blatantly
> not linear algorithms, but I'm only profiling the basic functions
> right now.
>
>
>
---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
>
__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Feb 3 16:16:12 2003