Re: 4 disks. Best utilization?

From: Joel Garry <joelga_at_rossinc.com>
Date: 1996/11/19
Message-ID: <1996Nov19.221129.9067_at_rossinc.com>#1/1


In article <32906B37.29CD_at_bulmer.com> Peter Wales <paw_at_bulmer.com> writes:
>Dana Stockler wrote:
>>
>> Soon we will be receiving two new computers to be used as Oracle
>> database
>> servers. Each computer will have 4 disks. What would be the best
>> placement
>> of the tablespace datafiles and the redo log files on these 4 disks?
>>
>> Each database will consist of the following tablespaces:
>> SYSTEM, <rollback segments>, <our tables>, <our indexes>, <temporary>.

Another corrollary to murphys law: When you have x number of things that should be on separate disks, you will have no more than x/2 disks.

Be sure that indexes and tables are on separate drives, and separate data tables that will be accessed simultaneously on separate drives. Separate out multiple rollback segments so that they toggle between or round-robin the disks - this allows the (likely) built-in disk caching to take load off the cpu i/o channels.

Don't forget your redo logs and control files. Put a control file on each disk. Put the redo logs on separate disks - whether you make them small or big depends on your app. And if you need archiving.

Modern disk controllers lessen the problem by re-ordering the actual writes, but you still can get some performance from separate controllers.

Buy several different books on tuning Oracle and read them. Whalen has a good one from Sams, Oracle press has a couple of good ones, Corrigan and Gurry from O'Reilly, all have the words "Oracle" and "Tuning" in the title.

You have enough disks for a fault-tolerant system, but it would be a dog.

>>
>> Any and all input will be welcomed.
>>
>> Regards,
>>
>> -----------------------------------------------------
>> Dana Stockler
>> -------------
>> Home: Solstadlia 10, 1364 Hvalstad, Norway
>> Work: IngeniørData as, Strøket 9, 1370 Asker, Norway
>> Email: stockler_at_ingdata.no
>> Phone: (+47) 66 78 97 41
>> -----------------------------------------------------
>
>Dana,
>you need to tell us more:
>How big is each disk?
>Are the disks on separate controllers?
>How big will each tablespace be?
>How many users?
>What type of application?
>also, what is the importance of down time? (you don't really have a
>enough disks for a failure-proof system).

Yes, and the most important determinant is the construction of the application itself.

>
>--
>+---------------------------------------------------------------+
>| Peter Wales |
>| IT Operational Services |
>| HP Bulmer Ltd "The Worlds Largest Cider Maker" |
>| MailTo:PAW_at_bulmer.com http://www.bulmer.com |
>+---------------------------------------------------------------+

-- 
Joel Garry               joelga_at_rossinc.com               Compuserve 70661,1534
These are my opinions, not necessarily those of Ross Systems, Inc.   <> <>
%DCL-W-SOFTONEDGEDONTPUSH, Software On Edge - Don't Push.            \ V /
panic: ifree: freeing free inodes...                                   O
Received on Tue Nov 19 1996 - 00:00:00 CET

Original text of this message