In this blog post series, we are looking at noteworthy enhancements to the Oracle Multitenant option of the Oracle Database in 12.1.0.2. In the previous post, we looked at the new Zone Maps feature, which can help in I/O pruning of data. Attribute Clustering is another new feature in 12.1.0.2. This means clustering data in close proximity based on contents of certain columns, i.e. data that logically belongs together. Attribute clustering basically improves the effectiveness of zone maps. It also makes Exadata Storage Indexes and In-memory min/max pruning more effective. I/O reduction is seen when used in conjunction with all these cases. Attribute clustering is defined at the table level, such as when creating the table: Create table …. clustering [join clause] by linear|interleaved order (column list); You can also specify an optional join clause inside the clustering clause. As an example, this can be used for joining a dimension table with a fact table and clustering the fact table with interleaved ordering. This allows the database to skip values in dimension columns during table scans. To combine attribute clustering with the zone map feature, an example syntax is shown below: Create table …. clustering by linear order (column list) yes on load yes on data movement with materialized zonemap [(optional zonemap name)]; The “yes on load” and “yes on data movement” mean that the clustering takes place when data is loaded into the table or when data movement takes place. If you do not specify “yes on load” or “yes on data movement”, then clustering is not enforced automatically but the type of clustering you want is stored as metadata that defines the natural clustering of the table. This can be used later for zone map creation. When clustering is not enforced automatically, it must be enforced by the user during loads. Altering an existing table is also possible, such as in the following example: alter table … add clustering by interleaved order (column list) yes on load yes on data movement; This command will not cluster the existing data in the table. But because “yes on data movement” is specified, the data will be clustered when a data movement operation is performed via “alter table .. move”. After the data is clustered in this way, a zone map can be defined on the table by modifying the clustering: alter table .. modify clustering with materialized zonemap [(optional zonemap name)]; For more information on the new Attribute Clustering feature, including the syntax examples in detail, please refer to the documentation (Database Data warehousing Guide) here .
↧