Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> c.d.o.server -> Re: 10g RAC, nonRAC difficulty, cost comparison

Re: 10g RAC, nonRAC difficulty, cost comparison

From: Howard J. Rogers <>
Date: Tue, 14 Dec 2004 11:30:21 +1100
Message-ID: <41be341b$0$1084$> wrote:
> Hi,
> I've been asked to come up with a general answer to this question:
> Comapred to standard oracle,
> how difficult is 10g RAC to setup, and maintain?
> My answer:
> oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo
> On a scale of 1 to 10
> RAC would be a 6 or 7
> and nonRAC is about a 3.
> One problem with RAC is the shared storage requirement.
> You will need to invest in an array and fibre channel technology.
> With nonRAC you can use cheap SCSI drives.
> If you are already investing in fibre channel,
> RAC is more compelling.
> oooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooo
> Do any of you Oracle experts have any opinions on this RAC
> nonRAC discussion?

Given that it's only an opinion, yes.

I really can't buy the 'if you've already got fibre channel, RAC is more compelling' argument. You do RAC because you need speed-up, scale-up, and/or high availability, not because you happen to have invested in a particular piece of hardware.

The fibre channel v. cheap SCSI argument also misses the less-obvious potential bottleneck in a true multi-instance RAC system: the speed, capacity and the latency of the cluster interconnect.

As a general rule (and this is only an opinion, remember) my advice is: unless you can come up with a truly compelling reason for RAC, then forget it. High availability is available these days in Data Guard, which is simpler and cheaper. Scale-up and speed-up might be achievable with modest new hardware purchases. Only if you can see these not meeting your requirements (eg, failover times in Data Guard are not good enough. Or the business is expanding at such a rate that another four CPUs is only going to buy us an additional year's grace), then go RAC.

It's those sorts of technical inputs that should determine nonRAC v. RAC purchasing decisions. Not "what make are your hard disks" type ones!

As to a specific answer to your specific question... Well, your question is a bit odd. I would assume that anybody about to install any flavour of Oracle would do their homework, and practice, and practice, and test, so that it should be a level 1 difficulty exercise by the time it is done for real in a production environment. And that would be as true for RAC for as it would be for non-RAC installations.

On the on-going maintenance side of things, my opinion is that RAC has more to go wrong (which is why I'd avoid it in the first place unless it also has commensurately more benefits for your organisation), and hence will require more maintenance. But that's quantity, not quality: the *difficulty* of the maintenance, for a trained RAC DBA, should not be greater.

Of course, if you just take someone who has done the odd Oracle 8i install (for example) and throw RAC at them, the difficulty scores will be somewhere in the stratosphere. But I'm assuming you wouldn't do that.

HJR Received on Mon Dec 13 2004 - 18:30:21 CST

Original text of this message