Re: [PATCH 0/7] [GSoC2009] Revision cache / git-daemon caching plan

From: Nicolas Pitre <nico@cam.org>
Date: Jun 5, 2009, 20:58 UTC
Message-ID: <alpine.LFD.2.00.0906051628530.3906@xanadu.home>
In-Reply-To: <m34ouu7h70.fsf@localhost.localdomain>

On Fri, 5 Jun 2009, Jakub Narebski wrote:

Show 8 quoted lines

> I like this sharing not only completed code, but plans, designs (and
> status reports) with Git Development Community (i.e. git mailing
> list).  I like this very much.
> 
> 
> I'd like to ask if there any results of profiling git server
> (git-daemon) code: how much is spend on object enumeration this GSoC
> project tries to make faster by the means of caching?

The git daemon only forks and execs other processes. It is hardly using any measurable CPU itself.

If you want to profile or simply have a good feel for what happens during a clone and clearly see what phase is problematic for the server then do this:

 1) cd to a local repository of your choice.  The bigger the better, 
    meaning that git itself is probably too small.  Try the linux 
    kernel, or a gcc mirror, or even better yet the gentoo repository.

 2) run this:

	git pack-objects --all-progress --revs --all --stdout \
		< /dev/null > /dev/null

 3) Sit and watch.  And for extra fun you may even measure the time 
    spent in each of the "counting objects", "compressing objects" and
    "writing objects" phases and compare them.

Nicolas