Re: Dev vs Prod: What is wrong here

meathammer wrote:
>> Aside from the obvious here is what you can not test:
>>
>> 1. Scalability
>> 2. Performance
>> 3. Memory interconnect related issues
>> 4. Load Balancing
>> 5. Failover
>> 6. Stability
>> 7. Resource utilization
>> 8. Interprocessor Parallel Query

> 
> good points thanks
> 
> Could you elaborate a bit on 
> #2 (I thought RAC was not about performance),
> #3
> #8
> 

#2: RAC is not about performance. But the performance on your test system will not equate to performance on your production system. Thus what runs well on one may be pathetically slow on the other.

#3: The single biggest issue I have seen in moving applications to RAC is poor architecture leading to issues with the memory interconnect. The interconnect is used to share information such as locks, blocks, and system change numbers. The so-called "spotlight effect" is legendary with RAC as it shines a bright light on issues created by poor designs leading to lots of block sharing. Your test system won't find them and they could be deadly.

#8: RAC uses IPQ to parallelize queries across nodes. Your test system will be incapable of doing this so what you see will not be what you will get: For better or worse.

-- 
Daniel A. Morgan
University of Washington
damorgan_at_x.washington.edu
(replace x with u to respond)
Puget Sound Oracle Users Group
www.psoug.org