|
Re: Oracle 10g finetuning for Hyper threaded CPUs [message #109791 is a reply to message #109636] |
Mon, 28 February 2005 22:46 |
pgongloo
Messages: 8 Registered: February 2005
|
Junior Member |
|
|
Oracle relies on the OS infrastructure for hyperthreading,it asks the OS how many CPU exist and in this case is returned the logical CPUs, and it tries to make maximum use of the available CPUs (assuming OS reports that hyperthreading is enabed).
Also, in general I guess it would depend on the application you're running. If the task(s) to be executed all have single serialization "choke" points, then Hyper threading won't help them, irrespective of Oracle.
If you're seeing a slowdown on thesystem, you can check the ADDM and AWR reports and figure out where the bottlenecks are for the period of interest.
-GP
|
|
|
Re: Oracle 10g finetuning for Hyper threaded CPUs [message #109796 is a reply to message #109791] |
Tue, 01 March 2005 00:19 |
gokhalej
Messages: 3 Registered: February 2005
|
Junior Member |
|
|
Hyperthreading enabled on the CPU, by default, makes Oracle recognise it as two processors, and automatically performs load balancing across the two virtual CPUs. I have read somewhere that this is indeed a deterrent to speed, expecially if block data is to ba handled as part of an update operation. Pls comment.
|
|
|
Re: Oracle 10g finetuning for Hyper threaded CPUs [message #109884 is a reply to message #109636] |
Tue, 01 March 2005 13:20 |
pgongloo
Messages: 8 Registered: February 2005
|
Junior Member |
|
|
In Oracle Database 10g, if for whatever reason you have block/segment contention, be it due to configuration of freelists or not using ASSM, or a hot block due to application issues or otherwise, if significant database time is spent on the block/segments, it will be reported by ADDM and with possible suggestions to fix it. It will tell you how much impact the operations were causing on your database.
In your case if there was a slowdown due to block contention, it will be pointed by the database. Can you fire up EM DB Console or use Command line interface to check for the ADDM findings when you run your load and reached steady state. You can also disable hyper-threading and perform the same procedure. and determine metrics (in ADDM/AWR report) that gauge the improvement or degradation of the system.
-GP
|
|
|