| RAC performance [message #177211] |
Tue, 13 June 2006 07:49  |
jlarchev
Messages: 3
Registered: June 2006
Location: Grenoble
|
Junior Member |
|
|
Hello
I have a 2 node cluster with 10gR2 on Linux RH AS 4
Everyday I have to refresh a schema with the production data exported using expdp
The 1st thing I do is to drop the schema to recreate an empty one.
This "drop user xxx cascade" takes 45 minutes.
Then the import (impdp) takes about 9 hours.
The last night before refreshing the database I stopped the instance running in the 2nd node.
The same "drop user xxx cascade" only took 7 minutes. And the impdp took 6 hours.
Can anybody explain that ?
|
|
|
|
| Re: RAC performance [message #177224 is a reply to message #177211] |
Tue, 13 June 2006 09:08  |
Frank Naude
Messages: 4587
Registered: April 1998
|
Senior Member |
|
|
Well, as you've just discovered - RAC is slower than a standalone instance. I've seen white papers claiming that certain operations can be up to 40% slower.
The more nodes you add to the cluster, the more overhead you add as Oracle needs to coordinate activities between all instances across the interconnect.
|
|
|
|
| Re: RAC performance [message #177227 is a reply to message #177224] |
Tue, 13 June 2006 09:23 |
jlarchev
Messages: 3
Registered: June 2006
Location: Grenoble
|
Junior Member |
|
|
Hi,
Thanks for your response.
>>Oracle needs to coordinate activities between all instances across the interconnect ??
What does this exactly means ?
|
|
|
|
| Re: RAC performance [message #179966 is a reply to message #177227] |
Thu, 29 June 2006 08:03 |
Frank Naude
Messages: 4587
Registered: April 1998
|
Senior Member |
|
|
|
Instances in the cluster need to talk to each other to keep things synchronized and prevent data corruption. Think data block transfers, lock mastering, etc.
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
