CHAPTER 21

Concurrency Control Techniques

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe

Introduction
 Concurrency control protocols
 Set of rules to guarantee serializability
 Two-phase locking protocols
 Lock data items to prevent concurrent access
 Timestamp
 Unique identifier for each transaction
 Multiversion currency control protocols
 Use multiple versions of a data item
 Validation or certification of a transaction

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 2

21.1 Two-Phase Locking Techniques
for Concurrency Control
 Lock
 Variable associated with a data item describing
status for operations that can be applied
 One lock for each item in the database
 Binary locks
 Two states (values)

Locked (1)
 Item cannot be accessed

Unlocked (0)
 Item can be accessed when requested

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 3

Two-Phase Locking Techniques
for Concurrency Control (cont’d.)
 Transaction requests access by issuing a
lock_item(X) operation

Figure 21.1 Lock and unlock operations for binary locks

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 4

Two-Phase Locking Techniques
for Concurrency Control (cont’d.)
 Lock table specifies items that have locks
 Lock manager subsystem
 Keeps track of and controls access to locks
 Rules enforced by lock manager module
 At most one transaction can hold the lock on an
item at a given time
 Binary locking too restrictive for database items

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 5

Two-Phase Locking Techniques
for Concurrency Control (cont’d.)
 Shared/exclusive or read/write locks
 Read operations on the same item are not
conflicting
 Must have exclusive lock to write
 Three locking operations

read_lock(X)

write_lock(X)

unlock(X)

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 6

Figure 21.2 Locking and
unlocking operations for
two-mode (read/write, or
shared/exclusive) locks

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21-7

Two-Phase Locking Techniques
for Concurrency Control (cont’d.)
 Lock conversion
 Transaction that already holds a lock allowed to
convert the lock from one state to another
 Upgrading
 Issue a read_lock operation then a write_lock
operation
 Downgrading
 Issue a read_lock operation after a write_lock
operation

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 8

Guaranteeing Serializability by Two-
Phase Locking
 Two-phase locking protocol
 All locking operations precede the first unlock
operation in the transaction
 Phases

Expanding (growing) phase
 New locks can be acquired but none can be released
 Lock conversion upgrades must be done during this phase

Shrinking phase
 Existing locks can be released but none can be acquired
 Downgrades must be done during this phase

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 9

Figure 21.3 Transactions that
do not obey two-phase
locking (a) Two transactions
T1 and T2 (b) Results of
possible serial schedules of
T1 and T2 (c) A
nonserializable schedule S
that uses locks

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 10

Guaranteeing Serializability by Two-
Phase Locking
 If every transaction in a schedule follows the two-
phase locking protocol, schedule guaranteed to
be serializable
 Two-phase locking may limit the amount of
concurrency that can occur in a schedule
 Some serializable schedules will be prohibited by
two-phase locking protocol

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 11

Variations of Two-Phase Locking
 Basic 2PL
 Technique described on previous slides
 Conservative (static) 2PL
 Requires a transaction to lock all the items it
accesses before the transaction begins

Predeclare read-set and write-set
 Deadlock-free protocol
 Strict 2PL
 Transaction does not release exclusive locks until
after it commits or aborts

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 12

Variations of Two-Phase Locking
(cont’d.)
 Rigorous 2PL
 Transaction does not release any locks until after it
commits or aborts
 Concurrency control subsystem responsible for
generating read_lock and write_lock requests
 Locking generally considered to have high
overhead

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 13

Dealing with Deadlock and Starvation
 Deadlock
 Occurs when each transaction T in a set is waiting
for some item locked by some other transaction T’
 Both transactions stuck in a waiting queue

Figure 21.5 Illustrating the deadlock problem (a) A partial schedule of T1′ and T2′ that is
in a state of deadlock (b) A wait-for graph for the partial schedule in (a)
Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 14
Dealing with Deadlock and Starvation
(cont’d.)
 Deadlock prevention protocols
 Every transaction locks all items it needs in
advance
 Ordering all items in the database

Transaction that needs several items will lock them
in that order
 Both approaches impractical
 Protocols based on a timestamp
 Wait-die
 Wound-wait

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 15

Dealing with Deadlock and Starvation
(cont’d.)
 No waiting algorithm
 If transaction unable to obtain a lock, immediately
aborted and restarted later
 Cautious waiting algorithm
 Deadlock-free
 Deadlock detection
 System checks to see if a state of deadlock exists
 Wait-for graph

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 16

Dealing with Deadlock and Starvation
(cont’d.)
 Victim selection
 Deciding which transaction to abort in case of
deadlock
 Timeouts
 If system waits longer than a predefined time, it
aborts the transaction
 Starvation
 Occurs if a transaction cannot proceed for an
indefinite period of time while other transactions
continue normally
 Solution: first-come-first-served queue
Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 17
21.2 Concurrency Control Based
on Timestamp Ordering
 Timestamp
 Unique identifier assigned by the DBMS to identify
a transaction
 Assigned in the order submitted
 Transaction start time
 Concurrency control techniques based on
timestamps do not use locks
 Deadlocks cannot occur

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 18

Concurrency Control Based
on Timestamp Ordering (cont’d.)
 Generating timestamps
 Counter incremented each time its value is
assigned to a transaction
 Current date/time value of the system clock

Ensure no two timestamps are generated during the
same tick of the clock
 General approach
 Enforce equivalent serial order on the transactions
based on their timestamps

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 19

Concurrency Control Based
on Timestamp Ordering (cont’d.)
 Timestamp ordering (TO)
 Allows interleaving of transaction operations
 Must ensure timestamp order is followed for each
pair of conflicting operations
 Each database item assigned two timestamp
values
 read_TS(X)
 write_TS(X)

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 20

Concurrency Control Based
on Timestamp Ordering (cont’d.)
 Basic TO algorithm
 If conflicting operations detected, later operation
rejected by aborting transaction that issued it
 Schedules produced guaranteed to be conflict
serializable
 Starvation may occur
 Strict TO algorithm
 Ensures schedules are both strict and conflict
serializable

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 21

Concurrency Control Based
on Timestamp Ordering (cont’d.)
 Thomas’s write rule
 Modification of basic TO algorithm
 Does not enforce conflict serializability
 Rejects fewer write operations by modifying
checks for write_item(X) operation

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 22

21.3 Multiversion Concurrency
Control Techniques
 Several versions of an item are kept by a system
 Some read operations that would be rejected in
other techniques can be accepted by reading an
older version of the item
 Maintains serializability
 More storage is needed
 Multiversion currency control scheme types
 Based on timestamp ordering
 Based on two-phase locking
 Validation and snapshot isolation techniques
Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 23
Multiversion Concurrency
Control Techniques (cont’d.)
 Multiversion technique based on timestamp
ordering
 Two timestamps associated with each version are
kept
 read_TS(Xi)
 write_TS(Xi)

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 24

Multiversion Concurrency
Control Techniques (cont’d.)
 Multiversion two-phase locking using certify locks

Three locking modes: read, write, and certify

Figure 21.6 Lock compatibility tables (a) Lock compatibility table for read/write
locking scheme (b) Lock compatibility table for read/write/certify locking scheme
Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 25
21.4 Validation (Optimistic) Techniques and
Snapshot Isolation Concurrency Control
 Optimistic techniques
 Also called validation or certification techniques
 No checking is done while the transaction is
executing
 Updates not applied directly to the database until
finished transaction is validated

All updates applied to local copies of data items
 Validation phase checks whether any of
transaction’s updates violate serializability

Transaction committed or aborted based on result

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 20- 26

Concurrency Control Based on
Snapshot Isolation
 Transaction sees data items based on committed
values of the items in the database snapshot
 Does not see updates that occur after transaction
starts
 Read operations do not require read locks
 Write operations require write locks
 Temporary version store keeps track of older
versions of updated items
 Variation: serializable snapshot isolation (SSI)

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 20- 27

21.5 Granularity of Data Items and
Multiple Granularity Locking
 Size of data items known as granularity
 Fine (small)
 Coarse (large)
 Larger the data item size, lower the degree of
concurrency permitted
 Example: entire disk block locked
 Smaller the data item size, more locks required
 Higher overhead
 Best item size depends on transaction type

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 20- 28

Multiple Granularity Level Locking
 Lock can be requested at any level

Figure 21.7 A granularity hierarchy for illustrating multiple granularity level locking

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 29

Multiple Granularity Level Locking
(cont’d.)
 Intention locks are needed
 Transaction indicates along the path from the root
to the desired node, what type of lock (shared or
exclusive) it will require from one of the node’s
descendants
 Intention lock types
 Intention-shared (IS)

Shared locks will be requested on a descendant
node
 Intention-exclusive (IX)

Exclusive locks will be requested
Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 30
Multiple Granularity Level Locking
(cont’d.)
 Intention lock types (cont’d.)
 Shared-intension-exclusive (SIX)

Current node is locked in shared mode but one or
more exclusive locks will be requested on a
descendant node

Figure 21.8 Lock compatibility matrix for multiple granularity locking

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 31

Multiple Granularity Level Locking
(cont’d.)
 Multiple granularity locking (MGL) protocol rules

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 32

21.6 Using Locks for Concurrency
Control in Indexes
 Two-phase locking can be applied to B-tree and
B+ -tree indexes
 Nodes of an index correspond to disk pages
 Holding locks on index pages could cause
transaction blocking
 Other approaches must be used
 Conservative approach
 Lock the root node in exclusive mode and then
access the appropriate child node of the root

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 33

Using Locks for Concurrency
Control in Indexes (cont’d.)
 Optimistic approach
 Request and hold shared locks on nodes leading
to the leaf node, with exclusive lock on the leaf
 B-link tree approach
 Sibling nodes on the same level are linked at
every level
 Allows shared locks when requesting a page
 Requires lock be released before accessing the
child node

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 34

21.7 Other Concurrency Control
Issues
 Insertion
 When new data item is inserted, it cannot be
accessed until after operation is completed
 Deletion operation on the existing data item
 Write lock must be obtained before deletion
 Phantom problem
 Can occur when a new record being inserted
satisfies a condition that a set of records accessed
by another transaction must satisfy
 Record causing conflict not recognized by
concurrency control protocol
Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 35
Other Concurrency Control Issues
(cont’d.)
 Interactive transactions
 User can input a value of a data item to a
transaction T based on some value written to the
screen by transaction T′, which may not have
committed
 Solution approach: postpone output of
transactions to the screen until committed
 Latches
 Locks held for a short duration
 Do not follow usual concurrency control protocol

Copyright © 2016 Ramez Elmasri and Shamkant B. Navathe Slide 21- 36