Re: RAC and scalability

From: hpuxrac <johnbhurley_at_sbcglobal.net>
Date: Tue, 8 Apr 2008 07:50:44 -0700 (PDT)
Message-ID: <a263894a-ab04-40e5-a0da-214a01ba39d4@f63g2000hsf.googlegroups.com>


On Apr 7, 6:52 am, Andrea <netsecur..._at_tiscali.it> wrote:
> Hi all,
> i've need some advise for resolve my doubt on RAC, because i have to
> sizing RAC environment in my business.
> Have been shows me two different solutions of RAC:
>
> 1)  One node of Bl680c (until 4 CPU quad Core)
>
> 2)  Two node of Bl480c (only 2 CPU quad Core)
>
> The type of choice is a valuate for future scalability, first solution
> support vertical scalability (2 to 4 CPU) instead of second solution
> where scalability is only for horizontal (add a nodes).
> First solution is correct sure but is more expensive and i would like
> to know if his cost is justifiable, more cpu are effective benefit?
>
> thanks for your info
>
> bye
> Andrew

I would start by recommending that you read "You probably don't need RAC" by Moans Nogood.

That pretty much gives you most of the critical things to think about.

RAC certainly has overhead that causes scalability problems for some applications your mileage may vary. Generic testing of hardware that doesn't really test out how well your applications scale out in RAC is an exercise in wasting time.

An excellent point that Moans makes is that the very act of trying to get high availability ( implementing RAC ) often causes lower availability. Does this site and company have the critical mass of experienced people necessary to support RAC? The questions that you are asking make me think "not so much".

Try locating that paper by Moans and do your homework ... that is my recommendation. Received on Tue Apr 08 2008 - 09:50:44 CDT

Original text of this message