Re: Is it a good idea, dev/test/production on one box

From: DA Morgan <damorgan_at_psoug.org>
Date: Wed, 20 Jul 2005 08:17:09 -0700
Message-ID: <1121872641.966590@yasure>

NewGuyOnRAC wrote:

> It is Oracle 9202 on RedHat Linux, 4cpu Dell computer, powerEdge, something 
> like that.  Load on production is not heavy, development is not heavy 
> neither, they bought their software from vendor, there is no in house 
> programming, so it is not the load/resource/performance issue that concerns 
> me, I just do not like the idea of whole company, one box, something going 
> wrong with the box, everything breaks.
> 
> I dont think it is the hardware, it is additional Oracle cpu/licence that 
> concerns them.  Can we just buy the licence for one box with support, and 
> another box with licence, but no support, will that work?
> 
> 
> "DA Morgan" <damorgan_at_psoug.org> wrote in message 
> news:1121821873.139233_at_yasure...
>

>>Depends on factors you haven't mentioned.
>>
>>1. What is the hardware platform and operating system
>>2. Storage subsystem
>>3. What is the Oracle edition and version
>>4. What is the load on each of the 3 instances
>>5. What kind of development is going on
>>6. What is the SLA for the production database.
>>
>>So, for example, if they are on 8i running on Windows and the production
>>load is coming from the internet serving up dynamic web pages, and the
>>development is low-level C++, and management expects the server to have
>>a 99.999999+% uptime it isn't going to work.
>>
>>I mean why not one database instance and 3 schemas?
>>
>>Your advice is good and they would be wise to follow it. I can't imagine
>>how they can afford Oracle licenses and not a few hundred/thousand
>>dollars for more hardware. My Grid Control machine, for example, is a
>>used IBM ThinkPad obtained from eBay for hardly any money at all.
>>--
>>Daniel A. Morgan
>>http://www.psoug.org
>>damorgan_at_x.washington.edu
>>(replace x with u to respond)

Please do not top-post.

Your instincts as I said before are good and they should not be doing all of this on a single machine.

The one exception, the one place where it is reasonable to put them together is with a 2 node RAC cluster. Using a cluster means that if one node burns itself to the ground, the other remains functional.

But look at it from their standpoint. Suppose they buy 2 new machines, one for test and one for dev. Suppose the production machine becomes toast: They are still out of business.

A far better strategy would be to leave things where they are ... buy one backup production machine ... and then use DataGuard or some other method to create a stand-by in case they lose their main production hardware. If they lose dev and test ... likely they can survive a week. If they lose production that could be quite another matter.

-- 
Daniel A. Morgan
http://www.psoug.org
damorgan_at_x.washington.edu
(replace x with u to respond)

Received on Wed Jul 20 2005 - 10:17:09 CDT