Preparing for Exadata - Oracle Exadata Recipes: A Problem-Solution Approach - page 148

Initial raw capacity is 252 TB
Final DATA usable capacity is 66.91 TB
Final RECO usable capacity is 16.72 TB
Final DBFS_DG usable capacity is .36 TB
Macintosh-7:source jclarke$

How It Works
Each storage server in an Exadata Database Machine has 12 physical disks of either the 600 GB High Performance or
the 3 TB High Capacity flavor. Based on your drive type and whether you've purchased a Quarter Rack, Half Rack, or
Full Rack, you will have a specific and fixed raw capacity for your database storage. The choice of ASM redundancy,
combined with the percentage allocation you decide for DATA and RECO database storage, will determine your usable
ASM disk group capacity.

6-7. Planning Database and ASM Extent Sizes
Problem
You're migrating databases to Exadata or installing new databases on Oracle Exadata and wish to configure your
database and Oracle ASM extent sizes for optimal performance.

Solution
You should create your ASM disk groups with a 4 MB allocation unit size for all ASM disk groups, per Oracle
recommendations. This recommendation is based on the physical mechanics of the storage cell SAS disks and the
1 MB I/O block size that Cell Services software operates with.

■ recipe 9-2 discusses aSM disk group attributes on exadata. please refer to allocation unit size settings and
other disk group attributes in this recipe.

Note

For your databases, Oracle recommends establishing large, (8 MB or greater) extent size for your large extents and
a minimum 1 MB extent size for smaller extents, if possible. The 8 MB extent size complements a 4 MB ASM allocation
unit size, reduces the proliferation of small extents in your database, which ultimately can help with capacity
management and performance, as well as minimizes the potential for suboptimal reads. Using a minimum of 1 MB
extent sizes aligns with the nature of how Exadata Storage Servers store data in 1 MB storage regions and issue I/Os
in 1 MB chunks.
You should use locally managed tablespaces and, generally speaking, auto-allocate extents in the tablespace
definitions or provide same-sized uniform extent sizes. With auto-allocation in place for your tablespaces and a
minimum extent size of 1 MB, you will ensure that your I/O request will be aligned with the storage infrastructure's
capabilities, which will help you avoid performing too many I/O requests or too few.

How It Works
There are a couple of key concepts to understand about how Oracle ASM allocates extents and how physical I/O is
performed inside the Exadata Storage Servers. First, with ASM (on Exadata or not), a segment's extents are striped
across disks in an ASM disk group in units of the allocation unit size, or AU size. For example, if you specify a