[Linux-cluster] RE: GFS performance analysis

Leonardo Rodrigues de Mello leonardo.mello at planejamento.gov.br
Thu Feb 2 14:37:11 UTC 2006

Dear Anthony,

	The tests that we report in the last message, were done localy in the
machine (without gnbd), and without lock (just one machine) using module
lock_nolock. We believe the performance problem we faced is about gfs
structure, not lock, but now, we are updating the webpage and results
with GNBD and GFS will be available soon, may be today.

	just to remember the page:

	Our objective here is to implement one Distributed Raid in a network
enviroment using gfs as the top filesystem. I had done that with success
about two years ago using ext3 at top filesystem. It was one distributed
raid in a network enviroment of 1.5 TB distributed in 4 machine servers
and 1 client mounting that raid. 

	Lets see how gfs performs, in the next tests

.''`.   Leonardo Rodrigues de Mello
: :' :  Coordenador de Projetos em Cluster e Grid
`. `'   DSI/SLTI/MP
  `-    55 61 3313 1329

> Dear Leonardo,
> We are facing the same problem on our cluster,
> We are actually using the GULM lock system, and are hoping that by 
> switching to the DLM locking system, we will resolve this problem.
> Wich locking method are you using for GFS?
> I will be gratefull, if you could keep me informed of your upcomming 
> performance/solutions.
> Regards,

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20060202/a0eb52dd/attachment.sig>

More information about the Linux-cluster mailing list