Sunteți pe pagina 1din 25

Copyright 2005, Oracle. All rights reserved.

Data Warehouse Schemas


2-2 Copyright 2005, Oracle. All rights reserved.
Objectives
After completing this lesson, you should be able to do
the following:
Explain the difference between a star schema and
a snowflake schema
Describe a star transformation
Control initialization parameters relating to star
queries
2-3 Copyright 2005, Oracle. All rights reserved.
Star Schema Model

SALES

AMOUNT
COST
QUANTITY_SOLD
CHANNELS

CHANNEL_ID
CHANNEL_DESC
CHANNEL_CLASS
CUSTOMERS

CUST_ID
CUST_GENDER
CUST_CITY
TIMES

TIME_ID
DAY_NAME
CALENDAR_YEAR
PROMOTIONS

PROMO_ID
PROMO_NAME

Dimension tables
Dimension tables
Fact table
PRODUCTS

PROD_ID
PROD_NAME
PROD_DESC
2-5 Copyright 2005, Oracle. All rights reserved.
Snowflake Schema Model
Dimension table
Dimension table
Dimension table
Dimension table
Fact table
CHANNELS
TIMES PRODUCTS
SALES
(amount,cost,
quantity_sold)
SUPPLIERS
CUSTOMERS
COUNTRIES
PROMOTIONS
2-7 Copyright 2005, Oracle. All rights reserved.
Creating Dimensions
CREATE DIMENSION products_dim
LEVEL product IS
(products.prod_id)
LEVEL subcategory IS
(products.prod_subcategory)
LEVEL category IS
(products.prod_category) ...

CATEGORY

SUBCATEGORY
PRODUCT
2-8 Copyright 2005, Oracle. All rights reserved.
Creating Hierarchy Relationships






CREATE DIMENSION products_dim
...
HIERARCHY prod_rollup (
product CHILD OF
subcategory CHILD OF
category)
ATTRIBUTE product DETERMINES
(products.prod_name, products.prod_desc,
prod_weight_class, prod_unit_of_measure,
prod_pack_size, prod_status, prod_list_price,
prod_min_price)
ATTRIBUTE subcategory DETERMINES
(prod_subcategory, prod_subcategory_desc)
ATTRIBUTE category DETERMINES
(prod_category, prod_category_desc);
2-9 Copyright 2005, Oracle. All rights reserved.
Third Normal Form Schemas
CUSTOMERS
PRODUCTS
ORDERS
ORDER
ITEMS
2-10 Copyright 2005, Oracle. All rights reserved.
Tuning Star Queries
A bitmap index should be built on each of the
foreign key columns of the fact table or tables.
Set STAR_TRANSFORMATION_ENABLED to TRUE.
This enables an important optimizer feature for star
queries.
It is set to FALSE by default
for backward compatibility.
Analyze all corresponding
objects.
select .

2-11 Copyright 2005, Oracle. All rights reserved.
Star Transformation
Star transformation is an optimization technique
that implicitly rewrites the original star query.
A star query is processed in two phases:
The necessary rows are retrieved from the fact
table, creating a result set.
The result set is then joined to the dimension
tables.
Star transformation requires that there be a single-
column bitmap index on every join column of the
fact table.
2-12 Copyright 2005, Oracle. All rights reserved.
Contrasting Star Transformation and
Conventional Joins
Conventional joins:
Cause each involved dimension table to be
successively joined to the fact table
Cause costly row-culling operations for each
successive dimension table join
Employ resource-consuming hash joins for row
culling
Star transformation:
Joins all involved dimension tables before the fact
table is accessed
Produces a composite set of row IDs before
accessing the fact table
Employs more efficient Boolean join operations
2-13 Copyright 2005, Oracle. All rights reserved.
Star Transformation Hints
STAR_TRANSFORMATION hint: Use the best plan
containing a star transformation, if there is one.
FACT(<table_name>) hint: The hinted table
should be considered as the fact table in the
context of a star transformation query.
NO_FACT (<table_name>) hint: The hinted table
should not be considered as the fact table in the
context of a star transformation.
FACT and NO_FACT hints are useful for star
transformation queries containing more than one
fact table.
2-14 Copyright 2005, Oracle. All rights reserved.
Star Query: Example
SELECT ch.channel_class, c.cust_city,
t.calendar_quarter_desc,
SUM(s.amount_sold) sales_amount
FROM sales s,times t,customers c,channels ch
WHERE s.time_id = t.time_id AND
s.cust_id = c.cust_id AND
s.channel_id = ch.channel_id AND
c.cust_state_province = 'CA' AND
ch.channel_desc IN ('Internet','Catalog') AND
t.calendar_quarter_desc IN
('1999-Q1','1999-Q2')
GROUP BY ch.channel_class, c.cust_city,
t.calendar_quarter_desc;
2-15 Copyright 2005, Oracle. All rights reserved.
Star Transformation: Rewrite Example
SELECT s.amount_sold
FROM sales s
WHERE time_id IN (SELECT time_id
FROM times
WHERE calendar_quarter_desc
IN('1999-Q1','1999-Q2'))
AND cust_id IN (SELECT cust_id
FROM customers
WHERE cust_state_province =
'CA')
AND channel_id IN(SELECT channel_id
FROM channels
WHERE channel_desc IN
('Internet','Catalog'));
2-17 Copyright 2005, Oracle. All rights reserved.
Star Transformation Execution Plan
SORT GROUP BY
HASH JOIN
HASH JOIN
TABLE ACCESS BY INDEX ROWID SALES
BITMAP CONVERSION TO ROWIDS
BITMAP AND
BITMAP MERGE
BITMAP KEY ITERATION
BUFFER SORT
TABLE ACCESS FULL CHANNELS
BITMAP INDEX RANGE SCAN SALES_CHANNELS_BX
BITMAP MERGE
BITMAP KEY ITERATION
BUFFER SORT
TABLE ACCESS FULL TIMES
BITMAP INDEX RANGE SCAN SALES_TIMES_BX
TABLE ACCESS FULL CHANNELS
TABLE ACCESS FULL TIMES
2-18 Copyright 2005, Oracle. All rights reserved.
Retrieving Fact Rows from One Dimension
BITMAP
KEY
ITERATION
Dimension
table
access
Fact table
bitmap
access
BITMAP
MERGE
One bitmap
is
produced.
2-19 Copyright 2005, Oracle. All rights reserved.
Retrieving Fact Rows from All Dimensions

Multiple bitmaps
are
ANDed together.
Result set
MERGE i

BITMAP AND
MERGE 1 MERGE n
BITMAP
conversion
to row IDs
2-20 Copyright 2005, Oracle. All rights reserved.
Joining the Result Set with Dimensions
HASH JOIN
Dimension 1
table
access
Fact table
access
from
result set
Dimension n
table
access
HASH JOIN
Dimension i
table
access
HASH JOIN
2-21 Copyright 2005, Oracle. All rights reserved.
Star Transformation: Further Optimization
In a star transformation execution plan, dimension
tables are accessed twice: once for each phase.
This may be a performance issue with big
dimension tables, and bad selectivity.
If the cost is cheaper, a temporary table may be
created and used instead of accessing the same
dimension table twice.
Done automatically, transparent to the end-user
The following PLAN_TABLE columns show
temporary tables information:
OBJECT_NAME
OTHER
2-22 Copyright 2005, Oracle. All rights reserved.
Bitmap Index Storage Performance
For each existing
bitmap index
Set
COMPATIBLE
to 10.2.0.0
Severe
slowdown before raising
COMPATIBLE?
Rebuild
Slowdown
after raising
COMPATIBLE?
Yes
Yes
Create new
bitmap indexes
2-23 Copyright 2005, Oracle. All rights reserved.
Static Partition Pruning and Star Query
Pruning is determined at parse time.
In most cases, fact tables are never directly
filtered.
SELECT prod_subcategory, cust_city,
sum(amount_sold)
FROM sales s, products p, customers c
WHERE s.prod_id = p.prod_id AND
s.cust_id = c.cust_id AND
prod_category IN ('Men','Women') AND
cust_year_of_birth ='1968' AND
s.time_id BETWEEN
TO_DATE('12.01.1999','mm.dd.yyyy') AND
TO_DATE('12.31.1999','mm.dd.yyyy')
GROUP BY prod_subcategory,cust_city;
2-24 Copyright 2005, Oracle. All rights reserved.
Static Partition Pruning Plan: Example
SORT (GROUP BY)
HASH JOIN
HASH JOIN
TABLE ACCESS (BY LOCAL INDEX ROWID) OF 'SALES' [12]
BITMAP CONVERSION (TO ROWIDS)
BITMAP AND
BITMAP MERGE
BITMAP KEY ITERATION
TABLE ACCESS FULL OF 'CUSTOMERS'
BITMAP INDEX (RANGE SCAN) OF 'SALES_C_BX' [12]
BITMAP MERGE
BITMAP INDEX (RANGE SCAN) OF 'SALES_T_BX' [12]
BITMAP MERGE
BITMAP KEY ITERATION
TABLE ACCESS (FULL) OF 'PRODUCTS'
BITMAP INDEX (RANGE SCAN) OF 'SALES_P_BX' [12]
TABLE ACCESS (FULL) OF 'CUSTOMERS'
TABLE ACCESS (FULL) OF 'PRODUCTS'
2-25 Copyright 2005, Oracle. All rights reserved.
Dynamic Partition Pruning and Star Query
Pruning is determined at run time.
Commonly used query form:
SELECT prod_subcategory, sum(amount_sold)
FROM sales s, products p, times t
WHERE s.prod_id = p.prod_id AND
s.time_id = t.time_id AND
t.calendar_month_number = 12 AND
t.calendar_year = 1999 AND
prod_category IN ('Men','Women')
GROUP BY prod_subcategory;
2-26 Copyright 2005, Oracle. All rights reserved.
Dynamic Partition Pruning Determination
During execution, the list of eligible partitions is
computed.
The corresponding recursive SQL can be retrieved
from the OTHER column of PLAN_TABLE.
SELECT DISTINCT TBL$OR$IDX$PART$NUM(
"SALES",0,d#,p#,"TIME_ID")
FROM (SELECT /*+ SEMIJOIN_DRIVER */ T.TIME_ID
FROM TIMES T
WHERE T.CALENDAR_MONTH_NUMBER = 12 AND
T.CALENDAR_YEAR=1999)
ORDER BY 1;
2-27 Copyright 2005, Oracle. All rights reserved.
Summary
In this lesson, you should have learned how to:
Distinguish star schemas and snowflake schemas
Interpret a star transformation execution plan
Set the STAR_TRANSFORMATION initialization
parameter
Use dynamic partition pruning
2-28 Copyright 2005, Oracle. All rights reserved.
Practice 2: Overview

This practice covers the following topics:
Analyzing star transformation execution plans
Understanding temporary table optimization for
star transformation

S-ar putea să vă placă și