On Fri, Jul 1, 2011 at 6:45 AM, Phil Pinkerton <pcpinkerton_at_gmail.com> wrote:
> Are there recommended standards with regards to Repository size, number of
> users per Repository, what type of data is contained in a Repository?
I've seen no published standards. It's relatively sensitive to layout:
A thousand users who aren't very active or are almost entirely
read-only might use a central server for a small repo. A bulky repo
that takes over an hour to check out due to thousands of individual
files and a tendency to have lots and lots and lots of auto-generated
branches, and people refuse to pick and choose their branches for
download, is a separate problem.
> Any experience with performance issues in regards to running Subversion on
> VMware vs a Blade Server ?
I've stuck with virtualization for testing repos only. Disk I/O tends
to be a serious bottleneck. The worst drawback I've had is when
someone kept insisting, insisting that it was fine to keep the disks
at 90% full because it was "stable" and "we'd know if something would
add to it". But the filesystem would *fragment* as data churned on
there, and I had to make charts to convince him to let me back up and
rebuild the filesystem, then we went back around over that fight six
months later. (It was storing a huge web proxy cache: the data churn
was ridiculous.)
This is more likely to happen in virtualization due to limiting the
disk resources. Think hard about your back end disk storage. Let us
know if you use attached storage such as a NetApp or fibre channel,
I've got some notes on aligning those filesystems for virtualization,
which makes a huge performance improvement.
Received on 2011-07-01 15:42:32 CEST