[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Large repositories, binary files, using W2K3 server

From: Mark Phippard <MarkP_at_softlanding.com>
Date: 2005-03-22 15:53:36 CET

Mike Burr <meburr@gmail.com> wrote on 03/22/2005 09:43:36 AM:

> I'm deploying Subversion in an environment where all work will be done
> within a single repository and nearly all files will be binary files
> 10s of MB in size (graphics files mostly).
>
> I'm wondering if there's any potential problems with having a very
> large repository, say 10s or possibly even 100s of GB. I'm thinking of
> things like performance and database corruption. The OS should handle
> files up to 16TB. I've noticed that Berkley DB doesn't split out the
> database files even when they get to be very large. Is it possible
> that things will slow with time or that the database files will become
> "fragmented", requiring a lot of seeking?
>
> Is there any problem that I might be overlooking by using Subversion
> to control binary-only data? Obviously, I don't plan to do any merging
> of files. Rather I just want to be able to version project folders and
> generally have more structure, collaboration and control.
>
> This will all be on a Windows 2003 Server machine (no choice!). Given
> that I'm using the latest stable version of everything (svn,
> apache2+mod_auth_sspi) does anyone foresee problems using this OS in a
> production environment.

The biggest problem in this sort of environment is typically the Working
Copy. You are aware that your WC will have two copies of each of these
files?

I would recommend using the fsfs repository format as it is a bit smaller
than BDB and doesn't have the problem you mention of storing everything in
a single file.

Finally, you might find that performance is not great when committing
these files, even the first time, as Subversion will always do a complete
binary delta of the file before it sends it over the wire. With really
big files that do not compress well, this is often a big time waster as it
would be better to just send the file without the compression.

I do not wish to discourage you, I would just recommend you prototype this
before you commit to it. I do not recall the name of the product, and it
probably isn't free, but I recall people mentioning another version
control tool that was created for your specific type of environment.

Mark

PS - One thing you might try once Subversion 1.2 is released is to use the
DAV autoversioning feature. This might actually work better for this type
of environment as you could avoid the use of a WC and some of the issues
it brings. The disadvantage is that your commits are more generic (no log
message and only one file per commit). Another plus is that your users
are just saving a file to a folder, they do not even need to know it is
being versioned.

_____________________________________________________________________________
Scanned for SoftLanding Systems, Inc. by IBM Email Security Management Services powered by MessageLabs.
_____________________________________________________________________________

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Mar 22 15:56:19 2005

This is an archived mail posted to the Subversion Users mailing list.