On Tue, 2004-07-06 at 18:22, Yannick Lecaillez wrote:
> Le mar 06/07/2004 à 19:07, Alvaro Herrera a écrit :
> > On Tue, Jul 06, 2004 at 06:17:16PM +0200, Yannick Lecaillez wrote:
> >
> > > What need to do (understand, to devel) to allow several
> > > postgres instance running from several server to access to the
> > > same data (no replication at all) hosted on a SAN ?
> >
> > Clustered shared memory, cluster-wide spinlocks. And with decent
> > performance, while at it ...
> Perhaps could be interesting to look at cluster file system which
> seems to have same problems and find solution about locking (i.e
> OpenGFS).
>
http://opengfs.sourceforge.net/showdoc.php?docpath=cvsmirror/opengfs/docs/ogfs-locking&doctitle=Locking&docauthor=ben.m.cahill(at)intel.com
>
> Found on google a clustered shared memory (openMosix project)
> http://www.unixreview.com/documents/s=8989/ur0404l/
>
> I would have the pgsql-hackers genius for do that :) . I think its the
> only feature which force company to buy 50000$ Oracle licence ...
>
I would note that Oracle first released OPS on UNIX at 7.0.13, in 1993.
Major performance issues were not resolved until 9i emerged, almost 10
years later...
It won't take PostgreSQL 10 years, but its impossible now, as Peter
observes.
Best regards, Simon Riggs