mutexed tdb benchmarks

[Volker Lendecke]

Hi!

Just to give you some very early, a bit hacked-up
results....

vlendec at sn-devel-104:~/git/scratch/lib/tdb$ time bin/tdbtorture -n 250 -l 2000 -s 1 -H 1000
Testing with 250 processes, 2000 loops, 1000 hash_size, seed=1
OK
real    0m40.764s
user    0m1.800s
sys     14m41.140s
vlendec at sn-devel-104:~/git/scratch/lib/tdb$ time bin/tdbtorture -n 250 -l 2000 -s 1 -H 1000 -m
Testing with 250 processes, 2000 loops, 1000 hash_size, seed=1
OK
real    0m1.995s
user    0m2.520s
sys     0m17.140s

This is with my private tdb-branch. If you take a look you
will see that I've commented out anything related to
transactions and traverse which still take out fcntl locks.
Both work, but the result is a bit less impressive :-)

This should a good picture of what we can achieve for
locking.tdb and others where in practice we don't do
transactions or traverses.

So if someone is out for a puzzle: If you look in mutex.c,
the fake I have to use at this moment is to store the
mutexes in a separate file. I'm not sure mmap()ing the mutex
area that lives inside the tdb file a second time together
with the normal tdb area is actually a good idea. I would
like to keep up the single file, but maybe the mutex and
data areas need to be split so that both mmap areas don't
overlap. More testing pending.

Volker

-- 
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de

**********************************************************
visit us at CeBIT: March 5th - 9th 2013, hall 6, booth E15
all about SAMBA and verinice, firewalls, Linux and Windows
free tickets available via email here : cebit at sernet.com !
**********************************************************