Re: Oracle RAC for scalability or High Availability only

From: DA Morgan <damorgan_at_psoug.org>
Date: Tue, 07 Mar 2006 11:08:35 -0800
Message-ID: <1141758503.511344@yasure.drizzle.com>

JEDIDIAH wrote:

>> The entire point of RAC is to NOT to buy hardware with HA features.


> 

> 	It rather depends on the shops. For some shops, a collection

> of dells isn't going to cut it. They're needs are too large.

Nonsense. You are again totally missing the basic concepts and architecture around RAC.

If it works for Amazon.com exactly how much bigger are you thinking? Provide numbers and specific examples if you think they exist.

>> You have entirely missed the point of RAC. Not just by a few inches
>> but by a few tens of kilometers.
>>
>> RAC is not about dual power supplies and such. It is about low cost
>> hardware. If something goes wrong you want the server to fall over.
>> The other nodes in the cluster, load balanced, pick up the processing.


> 

> 	The beauty of NUMA hardware is it ALREADY DOES THAT and is

> remarkably simpler and better tested than similar RAC configurations.

And the absolute total stupidity of NUMA is that it does NOT do what RAC does and your ignorance of RAC will not compensate for this lack of capability.

NUMA is totally incapable of doing what RAC does. And if you've reason to think otherwise then fill in the check-boxes.

Built with low-cost commodity hardware
Transparent failover equivalent to (TAF and FCF)
FAN API
Load balancing
Parallelizing
Supported by the Grid Control for management ease

> You don't have to futz with stupid shell scripts in CRS_HOME when you
> want to add 4 more cpus, you just add a system board and dbas don't
> even have to be aware of it.

I haven't a single shell script running any of my RAC clusters. Do you actually have any real-world contact with RAC?

>> A typical 2 CPU node for a RAC cluster is less than $5000 USD. If you
>> are spending more than that you need to go find your Oracle salesperson
>> and check into a re-education camp.


> 

> 	Then add in the cost of special storage hardware and someone

> to manage it.

What special storage hardware? Perhaps you should look up RAW and ASM or OCFS2. Someone to manage it? I think I can handle a browser window Ok. It so far has not taxed my brain beyond the breaking point.

Then add in the exta overhead of managing N systems
> rather than one (more bodies).

Huh?

> Add in the extra cost of dbas and
> sysadmins that are capable of dealing with the new complexities of the
> various clusterware.

My 80+ year old mother could be trained to manage a RAC cluster. What is your commission spamming for NUMA?

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

Received on Tue Mar 07 2006 - 13:08:35 CST