I had the same problem when putting together a proof of concept for my
company 6 weeks ago. Most examples I see are open source with projects
smaller than the linux kernel.
I can tell you that my company is planning on switching over early this
summer from a proprietary product (its such a bad product, I wont
mention it). We have 250 developers (+100 testers) and just under 200k
files. Our HEAD checkout will be around 12GB for all our projects
combined. Its mostly a machine issue from what I have seen.
Our current plan is to use a raid10 (1TB) setup on a box with 4GB of RAM
and quad opertons running a 64 bit linux kernel. It will have four
network cards that will have particular projects/users assigned to them
so that the IO load is not too high. If that runs too slow we will
separate projects onto other boxes. In our performance tests (we setup
some bots to peg the system) a 2.4Ghz p4 with a 5400rpm drive kept up
with 50 developers just fine. It slows down, but its still respectable.
We dont anticipate any major problems wrt machine requirements. Another
point worth mentioning is that we also plan on using bdb to start.
Open source projects are not a very good gauge because they typically
dont peak as high. They may have more developers, but the peak IO is
much lower. A repository of the scope that we are talking about
requires some serious hardware no matter what product you use. If the
machine or product fails, we have ~350 people picking their noses and
screaming. We chose subversion because of its simplicity and its
potential for integration to other systems (CRM, support, tracking).
I will definitely post our results when the conversion is completed as I
know that many large companies are looking for a canary in the coal
mine. I hope to post back around May with our results.
Brad
On Wed, 2005-02-09 at 10:37 -0700, Jani Averbach wrote:
> On 2005-02-09 11:32-0600, Ben Collins-Sussman wrote:
> >
> > On Feb 9, 2005, at 11:20 AM, Jani Averbach wrote:
> > >
> > > I will repeat my question of some months ago: can anyone actually
> > > give an example of a repository with anywhere near as many files
> > > and revisions as ours, with a couple of hundred active developers
> > > who semi-regularly check in, and well over 100 simultaneous
> > > checkouts at peak periods, that is managed by Subversion?
> > >
> >
> > How about the ASF repository?
>
> Exactly, the point is that where a casual wanderer can find those
> facts easily? I think it is good service for our (potential) users if
> we list (again) these facts somewhere on our www-site.
>
> BR, Jani
>
Received on Wed Feb 9 23:31:03 2005