Tuning the Storage Subsystem - Expert Oracle RAC Performance Diagnostics and Tuning - page 370

The ARBn background process will perform datafile reorganization. The time taken to
complete the reorganization is directly dependent on the number of ARBn processes started
and, as discussed earlier, it depends on the value of the parameter ASM_POWER_LIMIT .

5.

6. When the RDBMS instance opens a file or when a new file is created by the database

administrator (DBA), the RDBMS instance interacts with the ASM instance as a client to
obtain the file layout from the ASM instance. It then performs the following actions: 8

a. The DBA issues a statement to create a tablespace or add a datafile; the database

requests ASM for file creation.

b. An ASM foreground process creates a Continuing Operation Directory (COD) entry

and allocates space for the new file across the diskgroup.

c. The ASMB background process on the database instance receives an extent map for

the new file from the ASM instance.

d. The file is now open and the database process initializes the file.

7. After initialization, the database process requests that the file creation is committed. This

causes the ASM foreground process to clear the COD entry and mark the file as created.

8. Acknowledgment of the file commit implicitly closes the file. The database instance will

need to reopen the file for future I/O.

9. Based on user activity on the database instance, any updates to the data on the ASM

devices are performed by the DBWR process on the RDBMS instance. Such activity is
performed using the layout obtained by the RDBMS instance from the ASM instance
(illustrated in Step 6 above).

Allocation Units
Within an ASM disk, space on a diskgroup is divided into allocation units (AU). The default AU size is one megabyte
and can be set to a different value during diskgroup creation. However, once the diskgroup is created, changing the
AU size would mean creating a new diskgroup with the new AU size and moving data from the existing diskgroup to
the new diskgroup. Larger AU sizes may be useful in very large database (VLDB) scenarios or when using specialized
storage hardware.
Space is allocated to ASM files in units known as file extents. The relationship between file extents and allocation
units is as follows. An extent contains:

• One AU for each of the first 20,000 extents

• Four AUs for each of the next 20,000 extents

• 16 AUs for extents above 40,000

For the first 20,000 extents within a file, each extent contains one allocation unit. Extents are allocated on different
disks in a round robin fashion in order to spread the file and distribute I/O during read/write operations. Beyond the
first 20,000 extents, space is allocated in groups of eight extents at a time. Each extent contains four allocation units for
the second 20,000 extents, or 16 allocation units for all extents beyond 40,000. Extents are still allocated on different
disks to spread the file out, but, in addition, each allocation unit is written in turn to a different extent until each set of
eight extents is filled. The result is that consecutive allocation units will always be in different extents.

8 Oracle Automatic Storage Management - Under-the-Hood & Practical Deployment Guide - Nitin Vengurlekar, Murali Vallath,
Rich Long., Oracle Press, 2008.