Tuning the Storage Subsystem - Expert Oracle RAC Performance Diagnostics and Tuning - page 373

The SGA is broken into the shared pool, large pool, and shared pool reserved sizes. Default values for these
parameters are:
NAME TYPE VALUE
------------------------------------ ----------- ------------
large_pool_size big integer 12M
max_dump_file_size string unlimited
sga_max_size big integer 272M
shared_pool_reserved_size big integer 9646899
shared_pool_size big integer 0
sort_area_size integer 65536
workarea_size_policy string AUTO
memory_max_target big integer 272M
memory_target big integer 272M
sga_target big integer 0
By default the SGA for the ASM instance is sized very small. In the majority of these situations, the default SGA
size is sufficient. However, when the application performs high I/O activity or when the ASM instance supports more
than six clients, adding additional resources to the ASM instance may be required to improve performance. For
example, increasing the LARGE_POOL_SIZE of the ASM instance helps in the communications between ASM and its
clients:
SELECT p.number_kfdpartner,
d.failgroup
FROM x$kfdpartner p,
v$asm_disk d
WHERE p.disk = 0
AND p.grp = 1
AND p.grp = group_number
AND p.number_kfdpartner = d.disk_number;
Data Dictionary Views to Monitor Storage
Oracle's method of providing visibility for the underlying metadata of the various features and database functionality
is to provide database views. ASM is no exception. In this section, two of the important views related to performance
are discussed.
V$ASM_OPERATION
In everyday business activity, additional storage space is required to persist data and to find additional storage space;
more disks need to be added to the existing diskgroups. One of the features of ASM is to balance I/O among all disks
in the diskgroup. This is possible because of the way ASM distributes data across the underlying disks in a diskgroup.
When disks are added or removed from an existing diskgroup, data is rebalanced or distributed across all the disks
equally. The V$ASM_OPERATION view helps monitor the rebalance operation.
While the redistribution of data across the available disks is an excellent feature, it's not without additional
overhead. The number of RAC instances impacts the diskgroup rebalance operation scalability. The speed of the
rebalance operation can be improved by increasing the number of ARBx processes with the init.ora parameter
ASM_POWER_LIMIT or with the POWER clause. Valid values for the ASM_POWER_LIMIT parameter and the POWER clause
are 1-11. Due to the global cache inter-instance activity and the coordination required for rebalancing, additional RAC
instances could cause a linear degradation in performance. One of the methods to force a quicker rebalance operation