Creating a File System Oracle Database Tips by Donald Burleson

Linux supports many different file systems. A file system is a specific arrangement of disk data which allows the operating system to name and retrieve files from disk. A file system can contain file attributes which control the way a file or directory may be accessed. In Linux, these controls take the form of ownership and permissions to read from, write to and execute given files. The dates and times of creation and modification can also be maintained on a file system.

Although Linux can read from and write to many different file systems, the Fedora Core 1 distribution uses the ext3 file system by default. Ext3 is already being used on the internal hard disks of each node. This file system is a journaled file system which means that as files are written to disk, the file system maintains its own consistency. Journaling allows for speedy recovery in the event of an unclean shutdown; therefore, restarting after a sudden loss of power does not require a lengthy consistency check of the file system.

The ext3 file system cannot be used to store the shared files of a clustered database.

Oracle Cluster File System

Oracle has created a file system specifically tailored to the needs of RAC database files called Oracle Cluster File Systemor OCFS for short. OCFS must be downloaded and installed as a separate program from the Oracle database. OCFS maintains voting data among all of the nodes. This data is used to keep a consistent image of each file on disk to the different nodes of the cluster.

Raw - Storing Database Files without using a File System

Creating a partition without a file system, and reading and writing to this partition is referred to as using a raw partition or a raw device or simply raw. It is possible to create database files or even an entire database using raw. There are advantages to using raw partitions, such as I/O performance. And although the increase in I/O over OCFS is generally less than 5%, raw partitions move larger buffers of input and output than those moved by file systems.

Raw partitions are generally more difficult to administer than files on a file system. For example, only one file can be created on a raw partition and its path and name will be lost to the path and name of a device such as ?/dev/raw/raw1.? Symbolic links can be used to point to the device, providing a more meaningful path and name on the file system. The operating system commands that can normally be used to copy (cp) or move (mv) a file cannot be used on a file stored on a raw partition.

Automatic Storage Management

Oracle's Automatic Storage Management(ASM) combines advantages of raw partitions with the advantages of file systems. Instead of formatting and mounting a file system, ASM duplicates and balances file extents across multiple disk partitions without the need for the administrator to decide where to put them.

Although this technology was created to make administering I/O easier, the administrator cannot view files from the operating system, a concept that may take some getting used to. As you will see in Chapter 11, Release 2 of 10g software includes a command line tool that allows ASM files to respond to a limited set of commands as if they were on a standard file system.

Other File Systems used to Store Shared Database Files

There are other file systems which could be used to store Oracle database files in a RAC cluster. The Network File Systemis one which is supported by Linux. There are also a number of vendor specific Cluster File Systems available for different varieties of UNIX.  However, they are not covered in this book.

If you want to learn RAC at home, get the bestselling book "Personal Oracle Real Application Clusters" by Edward Stoever. You can buy it direct from the publisher for 30%-off and get instant access to the code depot of Oracle tuning scripts.