Recovery File Structures   «Prev  Next»
Lesson 10 Physical file placement on disk
Objective Discuss organizing your files on disk

Physical file placement

It would be impossible for me to suggest guidelines that fit your particular data base configuration. I will discuss structuring your system to configure for high performance and availability with low maintenance. Some of the concepts discussed here reference The OFA Standard | Oracle for Open Systems (part number A19308) originally written by Cary Millsap. This document was written for Oracle7 Unix installations and introduces a structure that can be tailored for other operating systems. Let us consider a possible file placement of your control files, redo logs, and the SYSTEM, TEMP, RBS, TOOLS, and USER tablespaces as outlined in the OFA document:
control file     /u01/oradata/sab/control.ctl

control file     /u02/oradata/sab/control.ctl

control file     /u03/oradata/sab/control.ctl

redo log group 1    /u03/oradata/sab/redo0101.log 

redo log group 1    /u05/oradata/sab/redo0102.log 

redo log group 1    /u07/oradata/sab/redo0103.log 

redo log group 2    /u04/oradata/sab/redo0201.log 

redo log group 2    /u06/oradata/sab/redo0202.log 

redo log group 2    /u08/oradata/sab/redo0203.log 

SYSTEM        /u02/oradata/sab/system01.dbf 

TEMP          /u04/oradata/sab/temp01.dbf 

RBS          /u03/oradata/sab/rbs01.dbf 
TOOLS        /u07/oradata/sab/tools01.dbf 

USERS        /u07/oradata/sab/users01.dbf 

Unix mount points

e are using Unix mount points 01 through 07 to indicate where files are placed. If we consider each mount point as a separate disk, you will need seven drives before you even identify disk space for your application data or your archive log files. Plus we must consider space for Unix or the Oracle8 base code that might also reside on separate disk drives.
Let’s consider how you can remap this for a Windows NT environment:

control file    d:\oradata\sab\control.ctl

control file    e:\oradata\sab\control.ctl

redo log group 1   d:\oradata\sab\redo0101.log 

redo log group 1   e:\oradata\sab\redo0102.log 

redo log group 2  d:\oradata\sab\redo0201.log 

redo log group 2   e:\oradata\sab\redo0202.log 

SYSTEM      e:\oradata\sab\system01.dbf 

TEMP       f:\oradata\sab\temp01.dbf 

RBS       e:\oradata/sab/rbs01.dbf 

TOOLS      e:\oradata/sab/tools01.dbf 

USERS      f:\oradata/sab/users01.dbf

Disk Drives

I have restructured this to use two disk drives to spread the load. In this example, we might install Windows NT, the Oracle source code, and our archive log files on the C:\ drive. The purpose of this exercise is to think about spreading out your disk I/O, to minimize the effect of a media failure and to help promote a robust system. Based on your environment you might also consider the use of mirrored (RAID 1) and block-interleaved with distributed parity (RAID 5) disk systems to help maximize the availability of your media.
The next lesson is about the state and structure of databases.