| Scale up vs Scale out [message #271210] |
Sun, 30 September 2007 19:42  |
seethem
Messages: 41
Registered: September 2007
|
Member |
|
|
Currently we are running -->
* 2-node Oracle RAC 10gR2 (10.2.0.3.0) on AIX.
* 16G 2CPU Dual core per node (IBM p550)
* 6.5GB of memory is allocated to the (SGA + PGA) per node out of the 16G.
FACTS
=======
* IBM Tivoli Monitoring (ITM) tool continually show O/S memory and CPU being highly utilized. I am not too worried about the CPU averaging about 70-80% utilized.
At this point management is basically interested in scaling each node up as opposed to scaling out.
QUESTIONS
===========
Can you suggest areas to look for or things that the dba team can do to propose what the current memory or CPU state should be?
I am only interested in providing facts that would answer "the why we should scale up?" question.
Many thanks, in advance.
|
|
|
|
| Re: Scale up vs Scale out [message #271370 is a reply to message #271210] |
Mon, 01 October 2007 08:04  |
Frank Naude
Messages: 4581
Registered: April 1998
|
Senior Member |
|
|
Well, both methods (scale up & scale out) should solve your performance/capacity problem.
Some of the reasons why they might prefer to scale up:
* Not enough floor space to accommodate more servers
* Prefer managing fewer servers (less work)
* Cheaper - only buy the components (no need to buy another chassis)
* Interconnect already saturated
Etc.
|
|
|
|
| Re: Scale up vs Scale out [message #271905 is a reply to message #271210] |
Wed, 03 October 2007 06:15 |
seethem
Messages: 41
Registered: September 2007
|
Member |
|
|
Hi Frank,
What can I do to quantify the scaling up?
What can I get to say -- because the load is this, you need to get X GB more of RAM or Y CPUs more
Is there a study that can point me in the right direction that would say, if x,y,z are happening, chances are that your system is overloaded.
Thanks for your advise and patience.
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
