Storage Administration - Oracle Exadata Recipes: A Problem-Solution Approach - page 249

DATA_CM01 3 22 o/192.168.10.4/DATA_CD_08_cm01cel02
DATA_CM01 3 23 o/192.168.10.4/DATA_CD_03_cm01cel02
DATA_CM01 3 26 o/192.168.10.3/DATA_CD_04_cm01cel01
DATA_CM01 3 32 o/192.168.10.3/DATA_CD_11_cm01cel01
DATA_CM01 3 34 o/192.168.10.3/DATA_CD_08_cm01cel01
... Additional disks omitted

24 rows selected.

SQL>

This output is telling us the following:

• Each disk has eight partner disks

• The partner disks for each of the disks are unique and balanced across the remaining two

failure groups but not located inside the same failure group. To illustrate this, consider
primary disk number 3, o/192.168.10.5/DATA_CD_03_cm01cel03 . This resides on cell server
192.168.10.5 and its partner disks are split evenly between 192.168.10.3 and 192.168.10.4

How It Works
With Oracle ASM normal redundancy or high redundancy, Oracle mirrors extents in an ASM disk group to one or
more partner disks. When mirroring extents, ASM uses a design concept called Disk Partnership. Each disk in a
normal or high redundancy disk group has one or more partners, used to mirror extents from the primary disk. If you
lose a primary disk, as well as the partner disk where extents are mirrored, you'll experience loss of access to data.
Oracle ASM tracks partner disk relationships for every disk in a disk group inside the Partnership Status Table, or
PST. Oracle reserves AU #1 on every ASM disk for the PST. So the PST specifies the list of partner disks for each disk.
Oracle Exadata environments have a minimum of 36 physical disks in a Quarter Rack and since this qualifies
as an environment with “a lot of disks,” ASM will maintain eight partner disks for each primary disk in the PST.
This means that for every primary extent allocated, Oracle will select from any one of the partner disks to write the
mirrored extents. For ASM disk groups configured with normal redundancy, Oracle will create the mirror extent on
one of the eight partner disks. With high redundancy, two mirror extents will be created on different partner disks.
It's worth stating again—ASM mirrors extents to partner disks on a different failure group. On Exadata, there is
one failure group per storage cell.

9-6. Measuring ASM Extent Balance on Exadata
Problem
To ensure that I/O requests are optimally balanced and serviced across Exadata Storage Server disks, you wish to
show the distribution of ASM extents across these Exadata disks in your Exadata environment.

Solution
In this recipe, we will provide several SQL scripts that will enable you to measure your database extent balance across
Oracle ASM grid disks. On Exadata and on all Oracle ASM environments, a well-balanced distribution of extents is
important for both performance and capacity management reasons.