Oracle allows you to configure "Flash-Based Grid Disks" from the PCIe flash cards that reside in each Exadata storage cell. This allows you to create ASM disk groups on flash storage and in theory, should yield solid-state performance gains for the segments residing in these ASM disk groups.

 

Oracle recommends a 4MB ASM allocation unit (AU) size for its ASM disk groups. In this post, we'll explore test cases with different ASM disk group AU sizes. Let's create some tablespaces to match our testing requirements:

 

In this post we'll perform tests for full-scanning a table when it's stored in an ASM disk group with CELL.SMART_SCAN_CAPABLE=FALSE and compare with the same table stored in a tablespace residing on an ASM disk group with CELL.SMART_SCAN_CAPABLE=TRUE. Let's create an ASM disk group to match this requirement:

 
What We're Trying to Prove
When customers deploy a new Exadata Database Machine, it will come pre-seeded with RAC-enabled database on ASM storage spanning the Exadata storage cells. This is one by Oracle's ACS group during the installation, and we at Centroid worked alongside Oracle to build our configuration to meet the specs in our configuration template.

 

With Oracle Exadata, what SQL statements actually are eligible for offload processing? If you've found yourself puzzled by the documentation, test results or not quite putting two and two together, this post attempts to define and prove what types of SQL statements truly are eligible for offload processing, and measure results.

 

As documented, parallel query operations are offload eligible with Exadata Database Machine. Let's put this to the test.

 

With Exadata Smart Scan processing, Exadata is capable of offloading a large portion of a query's workload to the storage cells. Many claims exists out there that recommend "migrate to Exadata and drop all your indexes", but let's see how Smart Scan processing works with various types of index access through a series of tests.

 

How do you tell whether a SQL statement running on Exadata is or was cell offload-able? The view GV$SQL (or V$SQL) provides the information you need to measure this, and of course this is also available in AWR SQL statistics views as well. The table below provides information regarding offload characteristics of individual SQL statements:

 

The purpose of this post is to demonstrate Smart Scan processing on Centroid's Exadata X2-2 Quarter Rack. Smart Scan processing is one of the components of Exadata Cell Offload Processing whose goal is to offload processing to storage cells instead of performing block IO requests on the database tier. The goal of Smart Scan processing is to do all the "heavy lifting", IO-wise, in the storage cell and only send back data to the compute nodes that is actually needed and requested.

 

The purpose of this post is to demonstrate Smart Scan processing on Centroid's Exadata X2-2 Quarter Rack. Smart Scan processing is one of the components of Exadata Cell Offload Processing whose goal is to offload processing to storage cells instead of performing block IO requests on the database tier.