This action might not be possible to undo. Are you sure you want to continue?
In Oracle8i performance improvements were made using materialized views to store the resulting rows of queries. The benefits of this mechanism are still relevant, but a certain subset of the queries used in a data warehouse may benefit from the use of Bitmap Join Indexes. ● How It Works ● Creation ● Restrictions
How It Works
In a Bitmap Index, each distinct value for the specified column is associated with a bitmap where each bit represents a row in the table. A '1' means that row contains that value, a '0' means it doesn't. Bitmap Join Indexes extend this concept such that the index contains the data to support the join query, allowing the query to retrieve the data from the index rather than referencing the join tables. Since the information is compressed into a bitmap, the size of the resulting structure is significantly smaller than the corresponding materialized view.
The index is created with reference to the columns in the joined tables that will be used to support the query. In the following example an index is created where the SALES table is joined to the CUSTOMERS table: CREATE BITMAP INDEX cust_sales_bji ON sales(customers.state) FROM sales, customers WHERE sales.cust_id = customers.cust_id; Since the CUSTOMERS.STATE column is referenced in the ON clause of the index, queries on the SALES table that join to the CUSTOMERS table to retrieve the STATE column can do so without referencing the CUSTOMERS table. Instead the data is read from the bitmap join index: SELECT SUM(sales.dollar_amount) FROM sales, customer WHERE sales.cust_id = customer.cust_id AND customer.state = 'California'; When dealing with large datasets, this reduction in processing can be substantial.
Bitmap Join Indexes have the following restrictions: ● Parallel DML is currently only supported on the fact table. Parallel DML on one of the participating dimension tables will mark the index as unusable. ● Only one table can be updated concurrently by different transactions when using the bitmap join index. ● No table can appear twice in the join. ● You cannot create a bitmap join index on an index-organized table or a temporary table. ● The columns in the index must all be columns of the dimension tables. ● The dimension table join columns must be either primary key columns or have unique constraints. ● If a dimension table has composite primary key, each column in the primary key must be part of the join. Hope this helps. Regards Tim...
In a data warehouse, B-tree indexes should be used only for unique columns or other columns with very high cardinalities (that is, columns that are almost unique). The majority of indexes in a data warehouse should be bitmap indexes. Bitmap indexes are most effective for queries that contain multiple conditions in the WHERE clause. Rows that satisfy some, but not all, conditions are filtered out before the table itself is accessed. This improves response time, often dramatically Bitmap indexes on partitioned tables must be local indexes Bitmap join indexes A bitmap join index is a space efficient way of reducing the volume of data that must be joined by performing restrictions in advance. For each value in a column of a table, a bitmap join index stores the rowids of corresponding rows in one or more other tables. In a data warehousing environment, the join condition is an equi-inner join between the primary key column or columns of the dimension tables and the foreign key column or columns in the fact table. Specify NOLOGGING clauses on the create index statement. Bitmap join indexes for snow flake schema CREATE BITMAP INDEX sales_c_gender_p_cat_bjix ON sales(customers.cust_gender, products.prod_category) FROM sales, customers, products WHERE sales.cust_id = customers.cust_id AND sales.prod_id = products.prod_id LOCAL NOLOGGING; Parallel DML is currently only supported on the fact table. Parallel DML on one of the participating dimension tables will mark the index as unusable. ● Only one table can be updated concurrently by different transactions when using the bitmap join index. ● The columns in the index must all be columns of the dimension tables.
B-tree indexes are most commonly used in a data warehouse to index unique or near-unique keys. In many cases, it may not be necessary to index these columns in a data warehouse, because unique constraints can be maintained
without an index, and because typical data warehouse queries may not work better with such indexes. Bitmap indexes should be more common than B-tree indexes in most data warehouse environments. References Oracle datawarehousing guide.pdf
● ● ● ●
If a B*tree index is not an efficient mechanism for accessing data, it is unlikely to become more efficient simply because you convert it to a bitmap index. Bitmap indexes can usually be built quickly, and tend to be surprisingly small. The size of the bitmap index varies dramatically with the distribution of the data. Bitmap indexes are typically useful only for queries that can use several such indexes at once. Updates to bitmapped columns, and general insertion/deletion of data can cause serious lock contention.
Alternatives to Oracle Fine Grained Access Control Bitmap join indexes Donald K. Burleson Oracle has introduced a new method to create speed join queries against very large data warehouse tables. This new method is called the bitmap join index, and this new table access method required the creation of an index that performs the join at index creation time and creates a bitmap index of the keys that are used in the join. For our example, we will use a many-to-many relationship where we have parts and suppliers. Each part has many suppliers and each supplier provides many parts
In this example, the database has 200 types of parts and the suppliers provide parts in all 50 states. The idea behind a bitmap join index is to pre-join the low cardinality columns together, thereby making the overall join faster: To create a bitmap join index we issue the following SQL. Note the inclusion of the FROM and WHERE clauses inside the CREATE INDEX syntax.
create bitmap index part_suppliers_state on inventory( parts.part_type, supplier.state) from inventory i, parts p, supplier s where i.part_id = p.part_id and i.supplier_id = p.part_id;
While b-tree indexes are used in the standard junction records, we can improve the performance of Oracle queries where the predicates involve the low cardinality columns. For example, look at the query below where we want a list of all suppliers of pistons in North Carolina:
select supplier_name from parts natural join inventory natural join suppliers where part_type = ‘piston’ and state = ‘nc’ ;
Prior to Oracle, this query would require a nested loop join or hash join of all three tables. In Oracle, we can pre-join these tables based on the low cardinality columns. For queries that have additional criteria in the WHERE clause that does not appear in the bitmap join index, Oracle will be unable to use this index to service the query. While Oracle markets this new feature with great fanfare, the bitmap join index is only useful for table joins that involve low-cardinality columns (e.g. columns with less than 300 distinct values). Bitmap join indexes are also not useful for OLTP databases because of the high overhead associated with updating bitmap indexes.
Oracle claims that this indexing method results in more than 8x improvement in table joins in cases where all of the query data resides inside the index. However, this claim is dependent upon many factors, and the bitmap join is not a panacea. In many cases the traditional hash join or nested loop join may out-perform a bitmap join. Some limitations of the bitmap join index join include: The indexed columns must be of low cardinality – usually with less than 300 distinct values ● The query must not have any references in the WHERE clause to data columns that are not contained in the index. ● The overhead when updating bitmap join indexes is substantial. For practical use, bitmap join indexes are dropped and re-built each evening about the daily batch load jobs. Hence bitmap join indexes are only useful for Oracle data warehouses that remain read-only during the processing day.
In sum, bitmap join indexes will tremendously speed-up specific data warehouse queries, but at the expense of pre-joining the tables at bitmap index creation time. If you like Oracle tuning, you might enjoy my latest book “Oracle Tuning: The Definitive Reference” by Rampant TechPress. It’s only $41.95 (I don’t think it is right to charge a fortune for books!) and you can buy it right now at this link: http://www.rampant-books.com/book_2003_1_Oracle_sga.htm
This chapter describes aspects of managing clusters. It contains the following topics relating to the management of indexed clusters, clustered tables, and cluster indexes: Guidelines for Managing Clusters Creating Clusters Altering Clusters Dropping Clusters Viewing Information About Clusters See Also: ○ Chapter 19, "Managing Hash Clusters" for a description of another type of cluster: a hash cluster ○ Chapter 14, "Managing Space for Schema Objects" is recommended reading before attempting tasks described in this chapter
● ● ● ● ●
Guidelines for Managing Clusters A cluster provides an optional method of storing table data. A cluster is made up of a group of tables that share the same data blocks. The tables are grouped together because they share common columns and are often used together. For example, the emp and dept table share the deptno column. When you cluster the emp and dept tables (see Figure 181), Oracle physically stores all rows for each department from both the emp and dept tables in the same data blocks. Because clusters store related rows of different tables together in the same data blocks, properly used clusters offer two primary benefits:
Disk I/O is reduced and access time improves for joins of clustered tables.
The cluster key is the column, or group of columns, that the clustered tables have in common. You specify the columns of the cluster key when creating the cluster. You subsequently specify the same columns when creating every table added to the cluster. Each cluster key value is stored only once each in the cluster and the cluster index, no matter how many rows of different tables contain the value. Therefore, less storage might be required to store related table and index data in a cluster than is necessary in non-clustered table format. For example, in Figure 181, notice how each cluster key (each deptno) is stored just once for many rows that contain the same value in both the emp and dept tables.
After creating a cluster, you can create tables in the cluster. However, before any rows can be inserted into the clustered tables, a cluster index must be created. Using clusters does not affect the creation of additional indexes on the clustered tables; they can be created and dropped as usual. You should not use clusters for tables that are frequently accessed individually.
Figure 18-1 Clustered Table Data
Text description of the illustration admin021.gif The following sections describe guidelines to consider when managing clusters, and contains the following topics: Choose Appropriate Tables for the Cluster Choose Appropriate Columns for the Cluster Key Specify Data Block Space Use Specify the Space Required by an Average Cluster Key and Its Associated Rows Specify the Location of Each Cluster and Cluster Index Rows Estimate Cluster Size and Set Storage Parameters See Also: ○ Oracle9i Database Concepts for more information about
● ● ● ● ● ●
clusters Oracle9i Database Performance Tuning Guide and Reference for guidelines on when to use clusters
Choose Appropriate Tables for the Cluster
Use clusters for tables for which the following conditions are true: The tables are primarily queried--that is, tables that are not predominantly inserted into or updated. ● Records from the tables are frequently queried together or joined.
Choose Appropriate Columns for the Cluster Key
Choose cluster key columns carefully. If multiple columns are used in queries that join the tables, make the cluster key a composite key. In general, the characteristics that indicate a good cluster index are the same as those for any index. For information about characteristics of a good index, see "Guidelines for Managing Indexes". A good cluster key has enough unique values so that the group of rows corresponding to each key value fills approximately one data block. Having too few rows for each cluster key value can waste space and result in negligible performance gains. Cluster keys that are so specific that only a few rows share a common value can cause wasted space in blocks, unless a small SIZE was specified at cluster creation time (see "Specify the Space Required by an Average Cluster Key and Its Associated Rows"). Too many rows for each cluster key value can cause extra searching to find rows for that key. Cluster keys on values that are too general (for example, male and female) result in excessive searching and can result in worse performance than with no clustering. A cluster index cannot be unique or include a column defined as long.
Specify Data Block Space Use
By specifying the PCTFREE and PCTUSED parameters during the creation of a cluster, you can affect the space utilization and amount of space reserved for updates to the current rows in the data blocks of a cluster's data segment. PCTFREE and PCTUSED parameters specified for tables created in a cluster are ignored; clustered tables automatically use the settings specified for the cluster. See Also: "Managing Space in Data Blocks" for information about setting the PCTFREE and PCTUSED parameters
Specify the Space Required by an Average Cluster Key and Its Associated Rows
The CREATE CLUSTER statement has an optional argument, SIZE, which is the estimated number of bytes required by an average cluster key and its associated rows. Oracle uses the SIZE parameter when performing the following tasks:
Estimating the number of cluster keys (and associated rows) that can fit in a clustered data block ● Limiting the number of cluster keys placed in a clustered data block. This maximizes the storage efficiency of keys within a cluster.
does not limit the space that can be used by a given cluster key. For example, if is set such that two cluster keys can fit in one data block, any amount of the available data block space can still be used by either of the cluster keys.
By default, Oracle stores only one cluster key and its associated rows in each data block of the cluster's data segment. Although block size can vary from one operating system to the next, the rule of one key for each block is maintained as clustered tables are imported to other databases on other machines. If all the rows for a given cluster key value cannot fit in one block, the blocks are chained together to speed access to all the values with the given key. The cluster index points to the beginning of the chain of blocks, each of which contains the cluster key value and associated rows. If the cluster SIZE is such that more than one key fits in a block, blocks can belong to more than one chain.
Specify the Location of Each Cluster and Cluster Index Rows
If you have the proper privileges and tablespace quota, you can create a new cluster and the associated cluster index in any tablespace that is currently online. Always specify the TABLESPACE option in a CREATE CLUSTER/INDEX statement to identify the tablespace to store the new cluster or index. The cluster and its cluster index can be created in different tablespaces. In fact, creating a cluster and its index in different tablespaces that are stored on different storage devices allows table data and index data to be retrieved simultaneously with minimal disk contention.
Estimate Cluster Size and Set Storage Parameters
The following are benefits of estimating a cluster's size before creating it: You can use the combined estimated size of clusters, along with estimates for indexes, rollback segments, and redo log files, to determine the amount of disk space that is required to hold an intended database. From these estimates, you can make correct hardware purchases and other decisions. ● You can use the estimated size of an individual cluster to better manage the disk space that the cluster will use. When a cluster is created, you can set appropriate storage parameters and improve I/O performance of applications that use the cluster.
Whether or not you estimate table size before creation, you can explicitly set storage parameters when creating each non-clustered table. Any storage parameter that you do not explicitly set when creating or subsequently altering a table automatically uses the corresponding default storage parameter set for the tablespace in which the table resides. Clustered tables also automatically use the storage parameters of the cluster. Creating Clusters
To create a cluster in your schema, you must have the CREATE CLUSTER system privilege and a quota for the tablespace intended to contain the cluster or the UNLIMITED TABLESPACE system privilege. To create a cluster in another user's schema you must have the CREATE ANY CLUSTER system privilege, and the owner must have a quota for the tablespace intended to contain the cluster or the UNLIMITED TABLESPACE system privilege. You create a cluster using the CREATE CLUSTER statement. The following statement creates a cluster named emp_dept, which stores the emp and dept tables, clustered by the deptno column:
CREATE CLUSTER emp_dept (deptno NUMBER(3)) PCTUSED 80 PCTFREE 5 SIZE 600 TABLESPACE users STORAGE (INITIAL 200K NEXT 300K MINEXTENTS 2 MAXEXTENTS 20 PCTINCREASE 33);
If no INDEX keyword is specified, as is true in this example, an index cluster is created by default. You can also create a HASH cluster, when hash parameters (HASHKEYS, HASH IS, or SINGLE TABLE HASHKEYS) are specified. Hash clusters are described in Chapter 19, "Managing Hash Clusters". See Also: Oracle9i SQL Reference for a more complete description of syntax, restrictions, and authorizations required for the SQL statements presented in this chapter
Creating Clustered Tables
To create a table in a cluster, you must have either the CREATE TABLE or CREATE ANY TABLE system privilege. You do not need a tablespace quota or the UNLIMITED TABLESPACE system privilege to create a table in a cluster. You create a table in a cluster using the CREATE TABLE statement with the CLUSTER option. The emp and dept tables can be created in the emp_dept cluster using the following statements:
CREATE TABLE emp ( empno NUMBER(5) PRIMARY KEY, ename VARCHAR2(15) NOT NULL, . . . deptno NUMBER(3) REFERENCES dept) CLUSTER emp_dept (deptno); CREATE TABLE dept ( deptno NUMBER(3) PRIMARY KEY, . . . ) CLUSTER emp_dept (deptno);
Note: You can specify the schema for a clustered table in the CREATE TABLE statement. A clustered table can be in a different schema than the schema containing the cluster. Also, the names of the columns are not required to match, but their structure must match.
Creating Cluster Indexes
To create a cluster index, one of the following conditions must be true:
Your schema contains the cluster. You have the CREATE ANY INDEX system privilege.
In either case, you must also have either a quota for the tablespace intended to contain the cluster index, or the UNLIMITED TABLESPACE system privilege. A cluster index must be created before any rows can be inserted into any clustered table. The following statement creates a cluster index for the emp_dept cluster:
CREATE INDEX emp_dept_index ON CLUSTER emp_dept INITRANS 2 MAXTRANS 5 TABLESPACE users STORAGE (INITIAL 50K NEXT 50K MINEXTENTS 2 MAXEXTENTS 10 PCTINCREASE 33) PCTFREE 5;
The cluster index clause (ON CLUSTER) identifies the cluster, emp_dept, for which the cluster index is being created. The statement also explicitly specifies several storage settings for the cluster and cluster index. Altering Clusters To alter a cluster, your schema must contain the cluster or you must have the ALTER ANY CLUSTER system privilege. You can alter an existing cluster to change the following settings: Physical attributes (PCTFREE, PCTUSED, INITRANS, MAXTRANS, and storage characteristics) ● The average amount of space required to store all the rows for a cluster key value (SIZE) ● The default degree of parallelism
Additionally, you can explicitly allocate a new extent for the cluster, or deallocate any unused extents at the end of the cluster. Oracle dynamically allocates additional extents for the data segment of a cluster as required. In some circumstances, however, you might
want to explicitly allocate an additional extent for a cluster. For example, when using Oracle9i Real Application Clusters, you can allocate an extent of a cluster explicitly for a specific instance. You allocate a new extent for a cluster using the ALTER CLUSTER statement with the ALLOCATE EXTENT clause. When you alter data block space usage parameters (PCTFREE and PCTUSED) or the cluster size parameter (SIZE) of a cluster, the new settings apply to all data blocks used by the cluster, including blocks already allocated and blocks subsequently allocated for the cluster. Blocks already allocated for the table are reorganized when necessary (not immediately). When you alter the transaction entry settings (INITRANS and MAXTRANS) of a cluster, a new setting for INITRANS applies only to data blocks subsequently allocated for the cluster, while a new setting for MAXTRANS applies to all blocks (already and subsequently allocated blocks) of a cluster. The storage parameters INITIAL and MINEXTENTS cannot be altered. All new settings for the other storage parameters affect only extents subsequently allocated for the cluster. To alter a cluster, use the ALTER CLUSTER statement. The following statement alters the emp_dept cluster:
ALTER CLUSTER emp_dept PCTFREE 30 PCTUSED 60;
See Also: Oracle9i Real Application Clusters Administration for specific uses of the ALTER CLUSTER statement in an Oracle Real Application Clusters environment
Altering Clustered Tables
You can alter clustered tables using the ALTER TABLE statement. However, any data block space parameters, transaction entry parameters, or storage parameters you set in an ALTER TABLE statement for a clustered table generate an error message (ORA-01771, illegal option for a clustered table). Oracle uses the parameters of the cluster for all clustered tables. Therefore, you can use the ALTER TABLE statement only to add or modify columns, drop non-cluster key columns, or add, drop, enable, or disable integrity constraints or triggers for a clustered table. For information about altering tables, see "Altering Tables".
Altering Cluster Indexes
You alter cluster indexes exactly as you do other indexes. See "Altering Indexes". Note: When estimating the size of cluster indexes, remember that the index is on each cluster key, not the actual rows. Therefore, each key appears
only once in the index. Dropping Clusters A cluster can be dropped if the tables within the cluster are no longer necessary. When a cluster is dropped, so are the tables within the cluster and the corresponding cluster index. All extents belonging to both the cluster's data segment and the index segment of the cluster index are returned to the containing tablespace and become available for other segments within the tablespace. To drop a cluster that contains no tables, and its cluster index, use the DROP CLUSTER statement. For example, the following statement drops the empty cluster named emp_dept:
DROP CLUSTER emp_dept;
If the cluster contains one or more clustered tables and you intend to drop the tables as well, add the INCLUDING TABLES option of the DROP CLUSTER statement, as follows:
DROP CLUSTER emp_dept INCLUDING TABLES;
If the INCLUDING TABLES option is not included and the cluster contains tables, an error is returned. If one or more tables in a cluster contain primary or unique keys that are referenced by FOREIGN KEY constraints of tables outside the cluster, the cluster cannot be dropped unless the dependent FOREIGN KEY constraints are also dropped. This can be easily done using the CASCADE CONSTRAINTS option of the DROP CLUSTER statement, as shown in the following example:
DROP CLUSTER emp_dept INCLUDING TABLES CASCADE CONSTRAINTS;
Oracle returns an error if you do not use the CASCADE CONSTRAINTS option and constraints exist.
Dropping Clustered Tables
To drop a cluster, your schema must contain the cluster or you must have the DROP ANY CLUSTER system privilege. You do not need additional privileges to drop a cluster that contains tables, even if the clustered tables are not owned by the owner of the cluster. Clustered tables can be dropped individually without affecting the table's cluster, other clustered tables, or the cluster index. A clustered table is dropped just as a non-clustered table is dropped--with the DROP TABLE statement. See "Dropping Tables". Note: When you drop a single table from a cluster, Oracle deletes each row of the table individually. To maximize efficiency when you intend to drop an entire cluster, drop the cluster including all tables by using the DROP
statement with the INCLUDING TABLES option. Drop an individual table from a cluster (using the DROP TABLE statement) only if you want the rest of the cluster to remain.
Dropping Cluster Indexes
A cluster index can be dropped without affecting the cluster or its clustered tables. However, clustered tables cannot be used if there is no cluster index; you must re-create the cluster index to allow access to the cluster. Cluster indexes are sometimes dropped as part of the procedure to rebuild a fragmented cluster index. For information about dropping an index, see "Dropping Indexes". Viewing Information About Clusters The following views display information about clusters: View
DBA_CLUSTERS ALL_CLUSTERS USER_CLUSTERS
Description view describes all clusters in the database. ALL view describes all clusters accessible to the user. USER view is restricted to clusters owned by the user. Some columns in these views contain statistics that are generated by the DBMS_STATS package or ANALYZE statement.
These views map table columns to cluster columns
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.