Session ID: 1542


When storage indexes kick in, the number of blocks that need to be returned from the storage cells drop dramatically. Since the storage cells can provide many blocks very quickly most of the time in big accesses is usually loading that information into the compute nodes. But storage indexes don't always kick in the way you would like and they have to be rebuilt by use in memory at every instance restart. Zonemaps have been created to eliminate the dependency on preceding use in a particular instance to prune block reading and delivery to the compute nodes. Physically ordering your data to match your zonemap definitions can minimize the number of blocks identified to be delivered. The nature of Data Warehouse loading, often with no updates other than periodic addition of data via loads, is synergistic with this new technology simply by making the newly loaded data conform to the physical order of the "keys" matching the zonemap definitions.

Objective 1: Understand how to define zonemaps usefully.

Objective 2: Understand how to load data with little extra overhead to keep zonemap references as small as possible.

Objective 3: Understand how the CBO chooses to use zonemaps in the plan.

Audience: In-Depth