1.5 - Data Con Currency and LEC

Data concurrency and locking
IBM Information Management Cloud Computing Center of Competence IBM Canada Labs
1
2011 IBM Corporation
Agenda
Transactions Concurrency & Locking Lock Wait Deadlocks
Supporting reading material & videos

Reading materials Getting started with DB2 Express-C eBook
Chapter 13: Concurrency and locking
Videos db2university.com course AA001EN

Lesson 8: Data concurrency and locking
Agenda
What is a transaction?
Your bank account
Balance = $1000
Your Moms bank account

Balance = $200
Transfer $100 from your account to your Moms account: - Debit $100 from your bank account (Subtract $100) - Credit $100 to your mom's bank account (Add $100)
What is a transaction? (cont'd)

One or more SQL statements altogether treated as one single unit Also known as a Unit of Work (UOW) A transaction starts with any SQL statement and ends with a COMMIT or ROLLBACK COMMIT statement makes changes permanent to the database ROLLBACK statement reverses changes COMMIT and ROLLBACK statements release all locks
Example of transactions
First SQL statement starts transaction INSERT INTO employee VALUES (100, 'JOHN') INSERT INTO employee VALUES (200, 'MANDY') COMMIT No changes applied due to ROLLBACK DELETE FROM employee WHERE name='MANDY' UPDATE employee SET empID=101 where name='JOHN' ROLLBACK
UPDATE employee SET name='JACK' where empID=100 COMMIT There is nothing to rollback ROLLBACK
7
Transactions ACID rules
Atomicity
All statements in the transaction are treated as a unit. If the transaction completes successfully, everything is committed If the transaction fails, everything done up to the point of failure is rolled back.
Consistency
Any transaction will take the data from one consistent state to another, so only valid consistent data is stored in the database
Isolation
Concurrent transactions cannot interfere with each other
Durability
Committed transactions have their changes persisted in the database
8
Agenda
Concurrency and Locking

App A
ID Name Peter John Mary Ann Age 33 23 22 55
App B App C App D
3 5 22 35
Concurrency: Multiple users accessing the same resources at the same time Locking: Mechanism to ensure data integrity and consistency
10
Locking
Locks are acquired automatically as needed to support a transaction based on isolation levels
COMMIT and ROLLBACK statements release all locks
Two basic types of locks:

Share locks (S locks) acquired when an application wants to read and prevent others from updating the same row Exclusive locks (X locks) acquired when an application updates, inserts, or deletes a row
11
Problems if there is no concurrency control

Lost update Uncommitted read Non-repeatable read Phantom read
12
Lost update
reservations
seat 7C 7B ...
name _____ _____
...
App A
App B
13
Lost update
reservations
seat 7C 7B ...
name _____ _____
...
App A update reservations set name = 'John' where seat = '7C'
App B
14
Lost update
reservations
seat 7C 7B ...
name _____ John _____
...
App B
15
Lost update
reservations
seat 7C 7B ...
name _____ John _____
...
App B update reservations set name = 'Mary' where seat = '7C'
16
Lost update
reservations
seat 7C 7B ...
name _____ Mary John _____
...
17
Lost update
reservations
seat 7C 7B ...
name _____ Mary John
...
? _____
18
Uncommitted read (also known as dirty read)

reservations
seat 7C 7B ...
App A
name _____ _____
...
App B
19

reservations
seat 7C 7B ...
name _____ _____
...
App B
20

reservations
seat 7C 7B ...
name _____ John _____
...
App B
21

reservations
seat 7C 7B ...
name _____ John _____
...
App B Select name from reservations where seat is '7C'
22

reservations
seat 7C 7B ...
name _____ John _____
...
John
23

reservations
seat 7C 7B ...
name _____ John _____
...
John
Roll back
24

reservations
seat 7C 7B ...
name _____ _____
...
John
Roll back
25

reservations
seat 7C 7B ...
name _____ _____
...

Further processing in App Further processing in App B uses incorrect / / B uses incorrect uncommitted value of uncommitted value of John John
John
Roll back
26
Non-repeatable read
reservations
seat 7C 7B ...
App A
name _____ _____
...
App B
27
Non-repeatable read
reservations
seat 7C 7B ...
App A
name _____ _____
...
App B
select seat from reservations where name is NULL
28
Non-repeatable read
reservations
seat 7C 7B ...
App A
7C 7B
name _____ _____
...
App B
29
Non-repeatable read
reservations
seat 7C 7B ...
App A
7C 7B
name _____ _____
...
App B update reservations set name = 'John' where seat = '7C'
30
Non-repeatable read
reservations
seat 7C 7B ...
App A
7C 7B
name _____ John _____
...
31
Non-repeatable read
reservations
seat 7C 7B ...
App A
7C 7B
name _____ John _____
...
select seat from reservations where name is NULL ... select seat from reservations where name is NULL
32
Non-repeatable read
reservations
seat 7C 7B ...
App A
7C 7B
name _____ John _____
...
7B
33
Non-repeatable read
reservations
seat 7C 7B ...
App A
7C 7B
name _____ John _____
...

The same SELECT (read) returns aa The same SELECT (read) returns different result: Less rows (in this different result: Less rows (in this case '7C' doesn't show anymore). case '7C' doesn't show anymore). This is aa non-repeatable read This is non-repeatable read
7B
34
Phantom read
reservations
seat 7C 7B ...
App A
name _____ Susan _____
...
App B
35
Phantom read
reservations
seat 7C 7B ...
App A
...
App B
36
Phantom read
reservations
seat 7C 7B ...
App A
7B
...
App B
37
Phantom read
reservations
seat 7C 7B ...
App A
7B
...
App B update reservations set name = NULL where seat = '7C'
38
Phantom read
reservations
seat 7C 7B ...
App A
7B
name _____ _____
...
39
Phantom read
reservations
seat 7C 7B ...
App A
7B
name _____ _____
...
40
Phantom read
reservations
seat 7C 7B ...
App A
7B
name _____ _____
...
7C 7B
41
Phantom read
reservations
seat 7C 7B ...
App A
7B
name _____ _____
...
7C 7B
42
The same SELECT (read) returns aa ... The same SELECT (read) returns different result: More rows (phantom select seat different result: More rows (phantom rows, in this case '7C', is shown) from reservations rows, in this case '7C', is shown) This is aa phantom read where name is NULL This is phantom read
Isolation levels
Policies to control when locks are taken DB2 provides different levels of protection to isolate data Uncommitted Read (UR) Cursor Stability (CS) Currently committed (CC) Read Stability (RS) Repeatable Read (RR)
43
Setting the isolation levels

Isolation level can be specified at many levels
Session (application), Connection, Statement For statement level, use the WITH {RR, RS, CS, UR} clause: SELECT COUNT(*) FROM tab1 WITH UR
For embedded SQL, the level is set at bind time For dynamic SQL, the level is set at run time
44
Comparing isolation levels
45
Cursor stability with currently committed (CC) semantics

Cursor stability with currently committed semantics is the default isolation level Use cur_commit db cfg parameter to enable/disable Avoids timeouts and deadlocks Cursor stability Cursor stability with currently committed
46

Cursor stability without currently committed
reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
Cursor stability with currently committed (Default behavior)

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
47

reservations
App A
update reservations set name = 'John' where seat = '7C'
seat
name Susan _____
...
App B
7C 7B ...

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
48

reservations
App A
seat
name Susan John _____
...
App B
7C 7B ...

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
49

reservations
App A
seat
...
App B S
select name from reservations where seat = '7C'
7C 7B ...

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
50

reservations
App A
seat
...
Lock Wait
App B
7C 7B ...

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
51

reservations
App A
seat
name
...
update select name Susan X 7C John S reservations from reservations 7B _____ set name = 'John' where seat = '7C' where seat = '7C' ... App B hangs until App A commits or rolls back which releases X lock
Lock Wait
App B

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
52

reservations
App A
seat
name
...
Lock Wait
App B

reservations
App A
53
seat
name Susan _____
...
App B
7C 7B ...

reservations
App A
seat
name
...
Lock Wait
App B

reservations
App A
54
seat
...
App B
7C 7B ...

reservations
App A
seat
name
...
Lock Wait
App B

reservations
App A
55
seat
...
App B S
7C 7B ...

reservations
App A
seat
name
...
Lock Wait
App B

reservations
App A
56
seat
...
App B S
7C 7B ...
App B retrieves currently committed value of 'Susan'

Comparing and choosing an isolation level

Isolation Level Repeatable Read (RR) ReadStability (RS) Cursor Stability (CS) Uncommitted Read (UR) Lost Dirty update Read Non-repeatable Phantom Read Read
Possible
Possible Possible
Possible Possible Possible
Application Type Read-write transactions Read-only transactions
High data stability required RS RS or RR
High data stability not required CS UR
57
Agenda
58
Lock wait
By default, an application waits indefinitely to obtain any needed locks

LOCKTIMEOUT (db cfg): Specifies the number of seconds to wait for a lock Default value is -1 or infinite wait
Example: (Same as when using isolation CS without CC):

reservations
App A
seat 7C 7B ...
name Susan _____
...
App B
59
Lock wait


reservations
App A
seat
name Susan _____
...
App B
7C 7B ...
60
Lock wait


reservations
App A
seat
...
App B
7C 7B ...
61
Lock wait


reservations
App A
seat
...
App B S
7C 7B ...
62
Lock wait


reservations
App A
seat
...
Lock Wait
App B
7C 7B ...
63
Lock wait


reservations
App A
seat
...
Lock Wait
App B
7C 7B ...
App B waits LOCKTIMEOUT seconds to get 'S' lock on first row

64
Agenda
65
Deadlocks

Occurs when two or more applications wait indefinitely for a resource Each application is holding a resource that the other needs Waiting is never resolved In the example, assume we are using isolation CS without CC
reservations
App A
seat 7C 7B ... 8E 9F
name Susan _____ Raul Jin
...
App B
66
Deadlocks

reservations
App A
seat
name Susan _____ Raul Jin
...
App B
7C 7B ... 8E 9F
67
Deadlocks

reservations
App A
seat
name Susan John _____ Raul Jin
...
App B
7C 7B ... 8E 9F
68
Deadlocks

reservations
App A
seat
name Susan John _____ Raul Jin
...
App B
update reservations set name = 'Sue' where seat = '9F'
7C 7B ... 8E 9F
69
Deadlocks

reservations
App A
seat
name Susan John _____ Raul Jin Sue
...
App B
7C 7B ... 8E 9F
70
Deadlocks

reservations
App A
update reservations set name = 'John' where seat = '7C' ... select name from reservations where seat = '9F'
71
seat
...
App B
7C 7B ... 8E
9F
Deadlocks

reservations
App A
72
seat
...
App B
7C 7B ... 8E
Lock Wait
9F
Deadlocks

reservations
App A
73
seat
...
App B S
update reservations set name = 'Sue' where seat = '9F' ... select name from reservations where seat = '7C'
7C 7B ... 8E
Lock Wait
9F
Deadlocks

reservations
App A
74
seat
...
7C 7B ... 8E
Lock Wait
App B
Lock Wait
9F
Deadlocks

reservations
App A
75
seat
...
7C 7B ... 8E
Lock Wait
App B
Lock Wait
9F
Deadlock!
Deadlocks
Deadlocks are commonly caused by poor application design DB2 provides a deadlock detector
Use DLCHKTIME (db cfg) to set the time interval for checking for deadlocks When a deadlock is detected, DB2 uses an internal algorithm to pick which transaction to roll back, and which one to continue. The transaction that is forced to roll back gets a SQL error. The rollback causes all of its locks to be released
76
Thank you!
77

1.5 - Data Con Currency and LEC

Uploaded by

Copyright:

Available Formats

You might also like

1.5 - Data Con Currency and LEC

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1.5 - Data Con Currency and LEC

Uploaded by

Copyright:

Available Formats

Data concurrency and locking

2011 IBM Corporation

Supporting reading material & videos

Videos db2university.com course AA001EN

2011 IBM Corporation

2011 IBM Corporation

Your Moms bank account

2011 IBM Corporation

What is a transaction? (cont'd)

2011 IBM Corporation

Transactions ACID rules

2011 IBM Corporation

Concurrency and Locking

App B App C App D

COMMIT and ROLLBACK statements release all locks

Two basic types of locks:

2011 IBM Corporation

Problems if there is no concurrency control

2011 IBM Corporation

name _____ _____

2011 IBM Corporation

name _____ _____

App A update reservations set name = 'John' where seat = '7C'

2011 IBM Corporation

name _____ John _____

App A update reservations set name = 'John' where seat = '7C'

2011 IBM Corporation

name _____ John _____

App A update reservations set name = 'John' where seat = '7C'

App B update reservations set name = 'Mary' where seat = '7C'

2011 IBM Corporation

name _____ Mary John _____

App A update reservations set name = 'John' where seat = '7C'

App B update reservations set name = 'Mary' where seat = '7C'

2011 IBM Corporation

name _____ Mary John

App A update reservations set name = 'John' where seat = '7C'

2011 IBM Corporation

Uncommitted read (also known as dirty read)

name _____ _____

2011 IBM Corporation

Uncommitted read (also known as dirty read)

name _____ _____

2011 IBM Corporation

Uncommitted read (also known as dirty read)

name _____ John _____

2011 IBM Corporation

Uncommitted read (also known as dirty read)

name _____ John _____

App B Select name from reservations where seat is '7C'

2011 IBM Corporation

Uncommitted read (also known as dirty read)

name _____ John _____

App B Select name from reservations where seat is '7C'

2011 IBM Corporation

Uncommitted read (also known as dirty read)

name _____ John _____

App B Select name from reservations where seat is '7C'

name _ _

name _ _

name _ John _

name _ John _

name _ Mary John _

name _ _

name _ _

name _ John _

name _ John _

name _ John _

name _ John _

name _ _

name _ _

name _ _

name _ _

name _ _

name _ _

name _ John _

name _ John _

name _ John _

name _ John _

name _ Susan _

name _ Susan _

name _ Susan _

name _ Susan _

name _ _

name _ _

name _ _

name _ _