Re: Oracle RAC for scalability or High Availability only

From: HansF <News.Hans_at_telus.net>
Date: Tue, 28 Feb 2006 19:52:47 GMT
Message-Id: <pan.2006.02.28.19.52.47.688991@telus.net>

On Tue, 28 Feb 2006 12:34:50 -0600, JEDIDIAH wrote:

> On 2006-02-28, HansF <News.Hans_at_telus.net> wrote:

>> On Mon, 27 Feb 2006 15:09:14 -0600, JEDIDIAH wrote:
>>
>>
>>> 	OTOH, adding nodes to the cluster is not just about the cost of 
>>> the boxes or of Oracle. Each box is going to your network and storage
>>> systems overhead. That next box might require you to get another high
>>> speed network switch (or 2 if you are being extra robust). The same goes
>>> with your fiber switches. Then there's the question of whether or not
>>> your storage array can handle that level of concurrency (nfs, iscsi or
>>> san).
>>
>> That argument cuts both ways.  If we're talking about would-a, could-a,
>> should-a and might-a scenarios, the opposite is also true - 
>>
>> your existing environment might have hit the limit on the box' network
>> access and you need to add the extra switch anyway; you might have spare


> 

> 	Except that a cluster typically has higher performance demands and 

> latency requirements. The network and storage gear required for RAC will
> more than likely be at least one notch above what you would need for the
> rest of your enterprise in general.
>
> You're not just buying into "more" you are also buying into
> inherently higher end equipment at the same time. Since it is an
> HA solution, you are also doubling everything.
>
> You're oversimplifying the costs involved and glossing over
> the none too trivial costs of SAN, switches, HBAs and robust GigE
> network gear. A robust nfs server is also no trivial matter either
> (if you choose to try and avoid SAN).
>

Don't blame all of the cost on Oracle. Or on Oracle RAC.

You are charging costs that belong to a HA infrastructure to the RAC costs center. Several of those costs would be required in any case to support HA, whether RAC or not. And a well developed HA infrastructure would often use the better quality NAS/SAN, the more robust storage, etc. whether RAC is there or not.

If the data is worth protecting, it is worth protecting well. You can get away with less robust equipment in a RAC environment, but it does beg the question - why bother with RAC if you aren't going to support it with reasonable equipment?

Or are you saying: Take any garbage NFS server for enterprise. Use a good one only when you use RAC. And then make sure everyone knows the only reason for getting the good (more expensive) one is due to RAC.

-- 
Hans Forbrich                           
Canada-wide Oracle training and consulting
mailto: Fuzzy.GreyBeard_at_gmail.com   
*** Top posting [replies] guarantees I won't respond. ***

Received on Tue Feb 28 2006 - 13:52:47 CST