Comparison testing: { FSFS, BDB } x { 1.5.4, trunk }

I did a little bit of simple comparison testing between FSFS and BDB in
1.5.4 and trunk. My testing involved loading a dumpfile of 5000 revisions
(taken from our own repository), then doing some single-revision dumps at a
few time-slices across the loaded repository.

Here are the highlights (percentages are ballpark estimates).

In trunk, FSFS:

  is significantly slower (30%) for writes operations. I have no idea why.

  is a bit faster for reads (20%).

  showed no meaningful disk usage changes. But I'm pretty sure this is
  an artifact of the testing dataset, which isn't as up-to-date-branch-heavy
  as more recent revision ranges in our source tree are.

In trunk, Berkeley DB:

  is significantly faster (50%) for write operations. This is almost
  certainly because post-commit deltification is doing a single
  deltification instead of touching a chain of files.

  is significantly slower (300%) for read operations. Distance to
  nearest fulltext?

  showed significant improvement in disk usage (20% savings) in trunk.
  For the same reasons that FSFS didn't show much improvement here, I
  must assume rep-sharing wasn't the real win here. More likely the
  minimization of fulltexts (one per line of history) is the win here.

In all things except disk usage (now in trunk), FSFS remains a clear winner
over BDB in this testing.

Attached are the script I used and a spreadsheet with the actual findings.

-- 
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand

#!/bin/sh

for svnadmin in \
    /home/cmpilato/projects/subversion/subversion/svnadmin/svnadmin \
    /home/cmpilato/projects/subversion-1.5.x/subversion/svnadmin/svnadmin ; do
    for fstype in fsfs bdb; do
        echo "### ${svnadmin} ${fstype}"
        for i in 1 2 3; do
            rm -rf repos
            ${svnadmin} create --fs-type ${fstype} repos
            echo "--- Dumpfile load"
            time -p ${svnadmin} load -q repos < dumpfile
            done
        echo "### --- Repository size"
        du -sk repos
        for rev in 10 2500 5000; do
            for i in 1 2 3; do
                echo "--- Revision dump (r${rev})"
                time -p ${svnadmin} dump -q repos -r ${rev} > /dev/null
                done
            done
        done
    done