Re: Hard disk configuration for fast write (Insert)

From: Volker Hetzer <firstname.lastname_at_ieee.org>
Date: Thu, 16 Nov 2006 16:06:01 +0100
Message-ID: <ejhuop$101$1@nntp.fujitsu-siemens.com>

Charles Hooper schrieb:
> Volker Hetzer wrote:

>> Sam schrieb:
>>> Hi There,
>>> What is the best Hard disk (Raid or striping) configuration (With minimum
>>> number of hard disk) for a Oracle database with lot's of
>>> Insert but not that much of read (query), also this database dos not need to
>>> be in archive mode (Data is not that critical right now).
>> Striping, typically 1MB stripes. No mirroring.
>> Of course, with a nologging database, the only recourse with a corrupt block
>> is to find out the relevant object and drop/recreate it. With a corrupt block
>> in a system table this means a reinstall.
>> But you /did/ ask for maximum write performance.

> A no logging database - an Oracle database without online redo logs?
> Would this Oracle database also not offer undo?
>
> The NOLOGGING clause applies to the initial creation of the objects
> only, not any DML that may take place on the tables, indexes, etc.
> after creation.
> http://asktom.oracle.com/pls/ask/f?p=4950:8:::::F4950_P8_DISPLAYID:9549144668497
The doc says different:
http://download-uk.oracle.com/docs/cd/B19306_01/server.102/b14200/clauses005.htm#CJAHABGF

But you are right, I mixed up the terminology. I meant, a database without archiving logs. For maximum performance, logging could be turned off too. It can be turned off at table level. A default can be set for the tablespace too. To be honest, thats what I'd recommend: Run the database in archive log mode and put the high performance data structures in a nologging tablespace. That way you can at least recover the system stuff if you screw up. Of course, system dependencies and triggers can still exist, negating the nologging benefits. Spatial is particularly nasty in that regard.

Lots of Greetings!
Volker

-- 
For email replies, please substitute the obvious.

Received on Thu Nov 16 2006 - 09:06:01 CST