Part3 Explain the Explain Plan

Master Product Manager
Oracle Database
February 2020
Maria Colgan
PART 3 INTERPRETING EXECUTION PLANS FOR SQL STATEMENTS
Explain the Explain Plan
@SQLMaria

Safe harbor statement
The following is intended to outline our general product direction. It is intended for information purposes
only, and may not be incorporated into any contract. It is not a commitment to deliver any material,
code, or functionality, and should not be relied upon in making purchasing decisions.
The development, release, timing, and pricing of any features or functionality described for Oracle’s
products may change and remains at the sole discretion of Oracle Corporation.
2
Confidential – © 2020 Oracle

What is an execution plan
Program Agenda
1
2
3
4
How to generate a plan
3
Understanding execution plans
Execution Plan Example
5
Partitioning
6 Parallel Execution

Program Agenda
1
2
3
4
4

What is an execution plan?
Query:
SELECT prod_category, avg(amount_sold)
FROM sales s, products p
WHERE p.prod_id = s.prod_id
GROUP BY prod_category;
HASH JOIN
TABLE
ACCESS
SALES
Tree-shaped representation of plan
TABLE
ACCESS
PRODUCTS
GROUP BY
Tabular representation of plan
5

Additional information under the execution plan
Access predicate
• Where clause predicate used for data
retrieval
• The start and stop keys for an index
• If rowids are passed to a table scan
6

Filter predicate
• Where clause predicate that is not used
for data retrieval but to eliminate
uninteresting row once
the data is found
7

Note Section
• Details on Optimizer features used such
as:
- Rule Based Optimizer (RBO)
- Dynamic Sampling
- Outlines
- SQL Profiles or plan baselines
- Adaptive Plans
- Hints (Starting in 19c)
8

Program Agenda
1
2
3
4
9

10
TKPROF
SQL Developer
SQL Monitor
Autotrace
Many ways to view an execution plan
…..But there are actually only 2 ways to generate one

1. EXPLAIN PLAN command
• Displays an execution plan for a SQL statement without actually executing the statement
2.V$SQL_PLAN
• A dictionary view introduced in Oracle 9i that shows the execution plan for a SQL statement that has been
compiled into a cursor in the cursor cache
Two methods for looking at the execution plan
How to generate an execution plan
Under certain conditions the plan shown with EXPLAIN PLAN can be
different from the plan shown usingV$SQL_PLAN
11

SQL> EXPLAIN PLAN FOR
SELECT p.prod_name, avg(s.amount_sold)
GROUP BY p.prod_name;
SQL> SELECT * FROM
table(dbms_xplan.display('plan_table',null,'basic'));
EXPLAIN PLAN command & dbms_xplan.display function
PLANTABLE
NAME
STATEMENT
ID
FORMAT
12

SQL> SELECT p.prod_name, avg(s.amount_sold)
GROUP BY p.prod_name;
SQL> SELECT * FROM
table(dbms_xplan.display_cursor(null, null, 'basic'));
Generate & display plan for last SQL statements executed in session
SQL_ID CHILD
NUMBER
FORMAT
• Format* is highly customizable - Basic ,Typical, All
– Additional low-level parameters show more detail
*More information on formatting on Optimizer blog
13

Program Agenda
1
2
3
4
• Cardinality
• Access paths
• Join methods
• Join order
14

What is it?
Estimate of number rows that will be returned by each operation
Cardinality
Why should you care?
It influences everything! Access method, Join type, Join Order etc.
How does the Optimizer Determine it?
Cardinality for a single column equality predicate = total num of rows
num of distinct values
For example: A table has 100 rows, a column has 5 distinct values =>
cardinality=20 rows
More complicated predicates have more complicated cardinality calculation
15

Identifying cardinality in an execution plan
Cardinality - estimated # of
rows returned
Determine correct cardinality using a SELECT
COUNT(*) from each table applying any WHERE
Clause predicates belonging to that table

SELECT /*+ gather_plan_statistics */
p.prod_name, SUM(s.quantity_sold)
WHERE s.prod_id =p.prod_id
GROUP BY p.prod_name ;
SELECT * FROM table (
DBMS_XPLAN.DISPLAY_CURSOR(FORMAT=>'ALLSTATS LAST'));
Checking cardinality estimates
17

Compare estimated number of rows (E-Rows) with actual
rows returned (A-Rows)
18

Extra information you get with ALLSTATS
Starts indicates the number of times that step, or operation was done
In this case the SALES table is partitioned and has 28 partitions
19

Buffers indicates the number of buffers that need to be read for each
step
20

0Mem - estimated amount of memory needed
1Mem - amount of memory needed to perform the operation in 1 pass
Used-Mem - actual amount of memory used and number of passes required
21

Checking cardinality estimates for Parallel Execution
Note: a lot of the data is zero in the
A-rows column because we only
show last execution of the cursor
which is done by the QC. Need to use
ALLSTATS ALL to see info on all
parallel server processes execution of
cursors
22

DBMS_XPLAN.DISPLAY_CURSOR(FORMAT=>'ALLSTATS ALL'));
Checking cardinality estimates for Parallel Execution
23

Easiest way to compare the estimated number of rows returned
with actual rows returned
Check cardinality using SQL Monitor
24

Solutions to incorrect cardinality estimates
Cause Solution
Stale or missing statistics DBMS_STATS
Data Skew Create a histogram
Multiple single column predicates on a table Create a column group using
DBMS_STATS.CREATE_EXTENDED_STATS
Function wrapped column Create statistics on the funct wrapped column
column using
DBMS_STATS.CREATE_EXTENDED_STATS
Multiple columns used in a join Create a column group on join columns using
using DBMS_STATS.CREATE_EXTENDED_STAT
Complicated expression containing columns
from multiple tables
Use dynamic sampling level 4 or higher
25

Program Agenda
1
2
3
4
• Cardinality
• Access paths
• Join methods
• Join order
26

Access Paths – Getting the data
Access Path Explanation
Full table scan Reads all rows from table & filters out those that do not meet the where clause predicates. Used when no index,
no index, DOP set etc.
Table access by Rowid Rowid specifies the datafile & data block containing the row and the location of the row in that block. Used if
Used if rowid supplied by index or directly in a where clause predicate
Index unique scan Only one row will be returned. Used when table contains a UNIQUE or a PRIMARY KEY constraint that
guarantees that only a single row is accessed e.g. equality predicate on PRIMARY KEY column
Index range scan Accesses adjacent index entries returns ROWID values Used with equality on non-unique indexes or range
range predicate on unique indexes (<.>, between etc)
Index skip scan Skips the leading edge (column) of the index & uses the rest Advantageous if there are few distinct values in the
values in the leading column and many distinct values in the non-leading column or columns of the index
index
Full index scan Processes all leaf blocks of an index, but only enough branch blocks to find 1st leaf block. Used when all
necessary columns are in index & order by clause matches index structure or if a sort merge join is done
Fast full index scan Scans all blocks in index used to replace a FullTable Scan when all necessary columns are in the index. Using
Using multi-block IO & can going parallel
Index joins Hash join of several indexes that together contain all the table columns that are referenced in the query. Won’t
Won’t eliminate a sort operation
Bitmap indexes Uses a bitmap for key values and a mapping function that converts each bit position to a rowid. Can efficiently
efficiently merge indexes that correspond to several conditions in a WHERE clause

Identifying access paths in an execution plan
If the wrong access method is being used check cardinality, join order…
Look in Operation section to see how an
object is being accessed
28

Table customers contains 10K rows & has a primary key on cust_id
SELECT country_id, name
FROM customers
WHERE cust_id IN (100,200,100000);
What plan would you expect for this query?
Access path example 1
Predicate Information (identified by operation id):
----------------------------------------------------------------------
3 - access("CUST_ID"=100 OR "CUST_ID"=200 OR "CUST_ID"=100000)
29

FROM customers
WHERE cust_id BETWEEN 100 AND 150;
30

FROM customers
WHERE country_name = ‘USA’;
31

Common access path issues
Issue Cause
Uses a table scan instead of index DOP on table but not index, value of MBRC
Picks wrong index Stale or missing statistics
Cost of full index access is cheaper than index look
look up followed by table access
Picks index that matches most # of column
32

Program Agenda
1
2
3
4
• Cardinality
• Access paths
• Join methods
• Join order
33

Join methods
Join Methods Explanation
Nested Loops joins For every row in the outer table, Oracle accesses all the rows in the inner table
Useful when joining small subsets of data and there is an efficient way to access the
second table (index look up)
Hash Joins The smaller of two tables is scan and resulting rows are used to build a hash table
on the join key in memory. The larger table is then scan, join column of the resulting
rows are hashed and the values used to probe the hash table to find the matching
rows. Useful for larger tables & if equality predicate
Sort Merge joins Consists of two steps:
1. Sort join operation: Both the inputs are sorted on the join key.
2. Merge join operation: The sorted lists are merged together.
Useful when the join condition between two tables is an inequality condition
34

Join types
Join Type Explanation
Inner Joins Returns all rows that satisfy the join condition
Outer Joins Returns all rows that satisfy the join condition and also returns all of the rows from the table
without the (+) for which no rows from the other table satisfy the join condition
Cartesian
Joins
Joins every row from one data source with every row from the other data source, creating
the Cartesian Product of the two sets. Only good if tables are very small. Only choice if
there is no join condition specified in query
Semi-Join Returns a row from the outer table when a matching row exists in the subquery data set.
Typically used when there is an EXISTS or an IN predicate, where we aren’t interested in
returning rows from the subquery but merely checking a match exists
Anti-Join Returns a row from the outer table when a matching row does not exist in the subquery data
set. Typically used when there is a NOT EXISTS or NOT IN predicate, where we aren’t
interested in returning rows from the subquery but merely checking a match doesn’t exists
A B

Identifying join methods in an execution plan
Look in the Operation section to check
the right join method is used
If wrong join type is used check stmt is written correctly &
cardinality estimates

SELECT e.last_name, e.salary, d.department_name
FROM hr.employees e, hr.departments d
WHERE d.departments_name IN ('Marketing‘,'Sales')
AND e.department_id = d.department_id;
Employees has 107 rows
Departments has 27 rows
Foreign key relationship between Employees and Departments on dept_id
What join method would you expect for this query?
Join method example 1

SELECT e.last_name, e.salary, d.department_name
FROM hr.employees e, hr.departments d
WHERE d.departments_name IN ('Marketing‘,'Sales')
AND e.department_id = d.department_id;

SELECT o.customer_id, l.unit_price * l.quantity
FROM oe.orders o, oe.order_items l
WHERE l.order_id = o.order_id;
Orders has 105 rows
Order Items has 665 rows

SELECT o.customer_id, l.unit_price * l.quantity
FROM oe.orders o, oe.order_items l
WHERE l.order_id = o.order_id;

SELECT o.order_id, o.order_date ,e.name
FROM oe.orders o , hr.employees e;
Orders has 105 rows
Employees has 107 rows

SELECT o.order_id, o.order_date ,e.name
FROM oe.orders o , hr.employees e;

SELECT s.quantity_sold
FROM sales s, customers c
WHERE s.cust_id =c.cust_id;
Sales table has 960 Rows
Customer table has 55,500 rows
Customer has a primary key created on cust_id
Sales has a foreign key created on cust_id

SELECT s.quantity_sold
WHERE s.cust_id =c.cust_id;
No join is needed
Table elimination transformation
Optimizer realizes that the join to
customers tables is redundant as no
columns are selected Presence of
primary –foreign key relationship
means we can remove table

What causes wrong join method to be selected
Issue Cause
Nested loop selected instead of hash join Cardinality estimate on the left side is
under estimated triggers Nested loop to
be selected
Hash join selected instead of nested loop In case of a hash join the Optimizer
doesn’t taken into consideration the
benefit of caching. If rows on the left
come in a clustered or ordered fashion
so the probe into 2nd table is more
efficient
Cartesian Joins Cardinality underestimation
Adaptive Plans in 12c can address
these problems on the fly by
changing the join method after
oracle sees what data is coming out
of the left-hand side of the join

Program Agenda
1
2
3
4
• Cardinality
• Access paths
• Join methods
• Join order
46

• The order in which the tables are join in a multi table statement
• Ideally start with the table that will eliminate the most rows
• Strongly affected by the access paths available
• Some basic rules
• Joins guaranteed to produce at most one row always go first
• Joins between two row sources that have only one row each
• When outer joins are used the table with the outer join operator must come after
the other table in the predicate
• If view merging is not possible all tables in the view will be joined before joining
to the tables outside the view
Join order

Identifying join methods in an execution plan
1
2
3
Want to start with the table that
reduce the result set the most
4
5
If the join order is not correct, check the statistics, cardinality &
access methods

It can be hard to determine Join Order forComplex SQL statements but
it is easily visible in the outline data of plan
SELECT * FROM table(dbms_xplan.display_cursor(format=>’TYPICAL +OUTLINE'));
Finding the join order for complex SQL
The leading hint tells
you the join order

What causes the wrong join order
Causes
Incorrect single table cardinality estimates
Incorrect join cardinality estimates

Program Agenda
1
2
3
4
• Cardinality
• Access paths
• Join methods
• Join order
51

SELECT e1.last_name, e1.job_title, e1.total_comp
FROM (SELECT e.manager_id, e.last_name, j.job_title,
e.salary+(e.salary+e.commission_pct) total_comp
FROM employees e, jobs j, departments d
WHERE d.department_name = 'Sales'
AND e.department_id = d.department_id
AND e.job_id = j.job_id ) e1,
(SELECT e.employee_id, e.salary+(e.salary+e.commission_pct) tc
FROM employees e, departments d
WHERE d.department_name = ‘Sales'
AND e.department_id = d.department_id ) e2
WHERE e1.manager_id = e2.employee_id
AND e1.total_comp >= e2.tc;
Find all the employees who make as much or more than their manager
Example SQL Statement

Is it a good execution plan?
Means no stats gathered strong
indicator this won’t be best
possible plan
1. Is the estimated number of rows
being returned accurate?
2. Are the cardinality
estimates accurate?
3.Are the access
method correct?

5. Is the join order correct? Is the table that
eliminates the most rows accessed first?
4. Are the right join methods being used?
1
2
3
4
5
Example cont’d execution plan

What does the plan tree look like?
TABLE ACCESS
EMPLOYEES
TABLE ACCESS
DEPARTMENT
MERGE JOIN
CARTESIAN
TABLE ACCESS
EMPLOYEES
HASH JOIN INDEX UNIQUE SCAN –
TABLE ACCESS DEPARTMENT
NESTED LOOP INDEX UNIQUE SCAN –
TABLE ACCESS JOBS
NESTED LOOP

Solution
2. Cardinalities are correct and with
each join number of rows reduced
1. Only 1 row is actually returned, and the cost is 4
lower now
4. Join methods have
changed to be all NL
3. Access methods
have changed for some
tables
1
2
3
5.The join
order has
changed
5
4

What does the plan tree look like?
TABLE ACCESS
DEPARTMENT
NESTED LOOP
INDEX UNIQUE SCAN –
TABLE ACCESS DEPARTMENT
NESTED LOOP
INDEX UNIQUE SCAN -
TABLE ACCESS JOBS
NESTED LOOP
NESTED LOOP
INDEX RANGE SCAN –
TABLE ACCESS EMPLOYEES
INDEX RANGE SCAN –
TABLE ACCESS EMPLOYEES

Program Agenda
1
2
3
4
58
5 Partitioning

59
Partitioning Provides Flexibility & Efficiency at Scale
Large Table
Difficult to
Manage
Partitions
Divide and Conquer
Easier to Manage
Improve Performance
Transparent to applications
SALES
JANUARY
MARCH
FEBRUARY
APRIL

SELECT SUM(s.sales_amount)
FROM sales s
WHERE s.sales_date
BETWEEN to_date('02/10/2020','MM/DD/YYYY')
AND to_date('02/10/2020','MM/DD/YYYY');
Only the 3 relevant partitions are accessed
What was the total sales for the weekend of Feb 10 - 12 2020?
Partition pruning
Feb 9th 2020
Feb 10th 2020
Feb 11th 2020
Sales Table
Feb 12th 2020
Feb 13th 2020
• Partition Pruning takes two forms
• Static pruning occurs when partitions are known in advance (via where clause)
• Dynamic pruning occurs when partitions are only known at runtime (via a join etc.)

Identifying partition pruning in a plan
If you see the word ‘KEY’ listed it indicate dynamic pruning
Pstart and Pstop list the partition
touched by the query

SELECT COUNT(*)
FROM RHP_TAB
WHERE CUST_ID = 9255 AND TIME_ID = '2020-02-10';
Why so many numbers in the Pstart / Pstop columns?
Numbering of partitions
Partition pruning

• An execution plan show
partition numbers for static
pruning
• Each partition is
numbered 1 to N
• Within each partition
subpartitions are
numbered 1 to M
• Each physical object in the
table is given an overall
partition number from 1 to
N*M
Partition pruning
Feb 5th 2020
RHP_TAB
Feb 10th 2020 Feb 15th 2020
PARTITION 1 PARTITION 5 PARTITION 10
….. …..
SUBPARTITION 1
SUBPARTITION 2
SUBPARTITION 1
SUBPARTITION 2
SUBPARTITION 1
SUBPARTITION 2
1
2
9
10
19
20

SELECT COUNT(*)
FROM RHP_TAB
WHERE CUST_ID = 9255 AND TIME_ID = '2020-02-10';
Why so many numbers in the Pstart / Pstop columns?
Partition pruning
Overall
partition #
Sub-
partition #
Range
partition #

• Advanced Pruning mechanism for complex queries
• Recursive statement evaluates the relevant partitions at runtime
• Look for the word ‘KEY’ in PSTART/PSTOP columns
Dynamic partition pruning
Partition pruning
SELECT sum(amount_sold)
FROM sales s, times t
WHERE t.time_id = s.time_id
AND t.calendar_month_desc IN
(‘JAN-20’,‘FEB-20’,‘MAR-20’);
Nov 2019
Dev 2019
Jan 2020
Sales Table
Feb 2020
Mar 2020
TIMES TABLE
• Dynamic partition pruning via a join

Dynamic partition pruning
Partition pruning

Identifying partition pruning in a plan
• What does :BF0000 mean?
Pstart and Pstop list the partition
touched by the query

Bloom Pruning
Hash Join
TRANSACTIONTABLE
1.Table scan:
ACCTS table is scanned and
matching rows to hash join
3.Table Scan:
Only partitions on theTRANS table
that have a 1 in the bit vector will
be scanned
4. Reduced row sent:
Only rows that match the
additional predicates from
partitions that match the
bit vector get sent hash
join
5. Hash Join: Join completed by probing into
the hash table from the trans acct_num’s to
find actual matching rows
2. Bloom Pruning list: A bit vector is
created that has a 1 for each absolute
partition there is a match for & 0 where
there is no match
Bloom Pruning list created
ACCTTABLE
Bit vector of partitions
00 01 1 1 1

Program Agenda
1
2
3
4
69
5 Partitioning

70
How it works
Parallel Execution
User connects
to the database
Background process
is spawned
Parallel servers
communicate among
themselves & the QC
using messages that are
passed via memory
buffers in the shared pool
QC gets parallel servers from
global pool and distributes
the work to them
Parallel servers - individual
sessions that perform work in
parallel Allocated from a pool of
globally available parallel server
processes & assigned to a given
operation
When user issues a parallel SQL
statement the background process
becomes the Query Coordinator

SELECT c.cust_last_name, s.time_id, s.amount_sold
WHERE s.cust_id = c.cust_id;
Steps completed by PX Coordinator versus Parallel Server Processes
Identifying parallel execution in the plan
Query Coordinator
Parallel Servers do majority of the work

• Data is divided into granules
• Block ranges
• Partition
• Each parallel server is allocated one
or more granules
• The granule method is specified on
the line above the scan operation in
the plan
How work is divided up among the parallel server processes
Identifying granules of parallelism in the plan

Access paths and how they are parallelized
Access Paths Parallelization method
Full table scan Block Iterator
Table accessed by Rowid Partition
Index unique scan Partition
Index range scan (descending) Partition
Index skip scan Partition
Full index scan Partition
Fast full index scan Block Iterator
Bitmap indexes (in StarTransformation) Block Iterator

Producers
ConsumersP1 P2 P3 P4
1. Hash join always
begins with a scan of
the smaller table. In
this case that’s is the
customer table. The 4
producers scan the
customer table and
send the resulting rows
to the consumers
P8
P7
P6
P5
WHERE c.cust_id = s.cust_id
Parallel Hash Join
How parallel execution works
Query
coordinator
CUSTOMERS
SALES
Hash Join with a Parallel Degree of 4

Producers
2. Once the 4
producers finish
scanning the customer
table, they start to scan
the Sales table and
send the resulting rows
to the consumers
P8
P7
P6
P5
Parallel Hash Join
Query
coordinator
CUSTOMERS
SALES
SALES

Producers
P8
P7
P6
P5
Parallel Hash Join
CUSTOMERS
SALES
3. Once the consumers receive
the rows from the SALES table
they begin to do the join. Once
completed they return the
results to the QC
Query
coordinator
SALES

Parallel Hash Join

Parallel Hash Join
TQ column shows parallel server sets

• Necessary when producers & consumers sets are used
• Producers must pass or distribute their data into consumers
• Operator into which the rows flow decides the distribution
• Distribution can be local or across other nodes in RAC
• Five common types of redistribution
Parallel distribution

• HASH
• Hash function applied to value of the join column
• Distribute to the consumer working on the corresponding hash partition
• Round Robin
• Randomly but evenly distributes the data among the consumers
• Broadcast
• The size of one of the result sets is small
• Sends a copy of the data to all consumers

• Range
• Typically used for parallel sort operations
• Individual parallel servers work on data ranges
• QC doesn’t sort just present the parallel server results in the correct order
• Partitioning Key Distribution – PART (KEY)
• Assumes that the target table is partitioned
• Partitions of the target tables are mapped to the parallel servers
• Producers will map each scanned row to a consumer based on partitioning column
• LOCAL suffix on the redistribution methods in a RAC database
• An optimization in RAC the rows are distributed to only the consumers
on the same RAC node

Indentifying parallel distribution in the plan
Shows how the PQ servers distribute rows between each other

Cardinality based distribution skew common scenario
• Crucial for parallel join of very small data sets with very large data sets
Distribution method decision based on expected number of rows
New adaptive distribution method HYBRID-HASH
• Statistic collectors inserted in front of PX process on the left hand side of the join
• If actual number of rows less than threshold, switch from HASH to Broadcast
- Threshold number of total rows < 2x DOP
Enabled by default
Hybrid-HASH Distribution method
Adaptive Distribution Method 1

Producers
P8
P7
P6
P5
Hybrid Parallel Hash Join
Adaptive Distribution Methods
Query
coordinator
DEPT
EMPS
Hybrid hash join
between DEPTS and
EMPS
Distribution method
based on runtime stats
Statistics collector
inserted in front of PX
processes scanning
the CUSTOMERS table
If # rows returned less
than threshold, rows
distributed via
Broadcast
Statistics Collector
threshold 2X DOP

• Hybrid hash join between EMP and
DEPT
• Distribution method based on
runtime stats
• Statistics collector inserted in front
of PX processes scanning DEPT

• If DEPT uses
BROADCAST
- EMP uses ROUND-
ROBIN
Broadcast/Round Robin
DOP used is 6
Number of rows returned is 4
Broadcast used because 24 rows distributed (6 X 4)

• If DEPT uses
BROADCAST
- EMP uses ROUND-ROBIN
• If DEPT used HASH
- EMP uses HASH
Broadcast/Round Robin
Hash/Hash
DOP used is 2
Hash used because only 4 rows distributed

Identifying parallel execution in a plan
IN-OUT column shows which step is run in parallel and if it is a single parallel server set or
not
PCWP - Parallel Combined With Parent - operation occurs when the database performs this step simultaneously
with the parent step
P->P - Parallel to Parallel – data is being sent from one parallel operation to another
P->S - Parallel to Serial - data is being sent to serial operation always happen on the step below the QC
NOTE If the line begins with an S then that step is executed serial – check DOP & access method

91
Related White Papers
• Explain the Explain Plan
• Understanding Optimizer Statistics
• Best Practices for Gathering Optimizer
Statistics
•What to expect from the Optimizer in 19c
•What to expect from the Optimizer in 12c
• What to expect from the Optimizer in 11g
Join the Conversation
https://twitter.com/SQLMaria
https://blogs.oracle.com/optimizer/
https://sqlmaria.com
https://www.facebook.com/SQLMaria

Part3 Explain the Explain Plan

Recommended

More Related Content

What's hot (20)

Similar to Part3 Explain the Explain Plan (20)

More from Maria Colgan (14)

Recently uploaded (20)

Part3 Explain the Explain Plan