[CTDB and GFS] test: scenario and some results

[Ignacio Coupeau]

simo wrote:
> On Fri, 2007-07-13 at 13:36 +0200, Ignacio Coupeau wrote:
>> GFS
>> --------------------------------
>> 1-node ops average = 7401 ops
>> 2-node ops average =  204 ops
>> 3-node ops average =  469 ops
> 
> Wow very bad numbers, what does the pingpong test give you on GFS?
> http://wiki.samba.org/index.php/Ping_pong
> 

for reference with ext3 local:
	./ping_pong c 4
		300850 locks/sec
	./ping_pong  -rw c 4
		202650 locks/sec


In all the test that follows, the release of the any node made the 
others speed up.

The nodes don't have any service running, only gfs et al.

1,2,3 nodes over the same file without options:
-----------------------------------------------
./ping_pong  /usr/local/etc/users_1/c 4

  1 node:  15897 locks/sec
  2 nodes:  1217 locks/sec
  3 nodes:    33 locks/sec


1,2,3 nodes over the same file with -rw options:
------------------------------------------------
./ping_pong -rw /usr/local/etc/users_1/c 4

1 node:  13708 locks/sec
2 nodes:   166 locks/sec
3 nodes:    11 locks/sec


1,2,3 nodes over the same file with -rwm options:
-----------------------------------------------
./ping_pong -rwm /usr/local/etc/users_1/c 4

1 node:  13599 locks/sec
2 nodes:   128 locks/sec
3 nodes:    14 locks/sec


As the data are so awful, I compiled the ping_pong in another gfs 
cluster with http in production (very few load, with similar hardware 
and RH4U3), the results are very similar.

Any idea(s)?
I have a very one: gfs (at least gfs1) is poor, very poor, for this issue.

The GPFS, runs fine?

Thanks,

Ignacio


-- 
________________________________________________________
Dr. Ignacio Coupeau
Systems and Network Services Director
IT Services
University of Navarra           http://www.unav.edu/
Pamplona, SPAIN                 http://www.unav.es/SI/