Re: Question about RAC

Hi rjamya

Not RAC expert here but I wonder if you achieve good performance sharing a couple interconnect (gigabit, infiniband?) between 55 RAC Databases? The bottleneck I have seen in some customers most of times are due to interconnect.

Rgds

On 10/31/06, rjamya <rjamya_at_gmail.com> wrote:
>
> Okay ... some basics ... we don't partition our servers .. so no lpar etc
>
> We run two node cluster, i.e. two physical servers in a cluster.
> separate boxes. For each RAC database, we have two instances, one and
> two. One instance of a RAC DB runs on one server and another on
> another server.
>
> 1. SERVER A and B host two RAC databases DB_A and DB_B.
> 2. SERVER A runs DB_A1 and DB_B1 instances
> 3. SERVER B runs DB_A2 and DB_B2 instances.
>
> Get the picture? Now step back and imagine instead of two databases
> across two nodes (that is two instances per node) we run 55 instances
> per node, that is 55 databases on the two node cluster.
>
> Yes, they connect to same san, use same pair of interconnects and are
> a mix of 9204, 9205, 10104 and 10202 databases.
>
> to answer your last question, yes
> xxx_at_yyy> ps -ef | grep pmon | wc -l gives me 50 and I know for a fact
> that at-elast 4 instances are down because they are being refreshed
> from production.
>
> Hope I make it clear.
> Raj
>
> On 10/31/06, amonte <ax.mount_at_gmail.com> wrote:
> > rjamya
> >
> > there wont be vpart, lpart and such, lets make a bit easier, imagine you
> > have two small PCs running 2 instance each, 4 instances in total and
> > supporting 2 RAC Databases, 4 instances shares Memory, CPU,
> Interconnect. Is
> > is logical?
> >
> > you said you run 55 RAC in two nodes but then you say you have a
> instance
> > per node....?!
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

--
http://www.freelists.org/webpage/oracle-l