Chapter 15Chapter 15Transaction Management

Outline Outline

Transaction basicsConcurrency controlRecovery managementTransaction design issuesWorkflow management

Transaction DefinitionTransaction DefinitionSupports daily operations of an

organizationCollection of database operationsReliably and efficiently processed as one

unit of workNo lost data

– Interference among multiple users– Failures

Airline Transaction ExampleAirline Transaction ExampleSTART TRANSACTION

Display greetingGet reservation preferences from userSELECT departure and return flight recordsIf reservation is acceptable then

UPDATE seats remaining of departure flight recordUPDATE seats remaining of return flight recordINSERT reservation recordPrint ticket if requested

End IfOn Error: ROLLBACK

COMMIT

Transaction PropertiesTransaction PropertiesAtomic: all or nothingConsistent: database must be consistent

before and after a transactionIsolated: no unwanted interference from

other usersDurable: database changes are permanent

after the transaction completes

Transaction Processing ServicesTransaction Processing ServicesConcurrency controlRecovery managementService characteristics

– Transparent– Consume significant resources– Significant cost component

Concurrency ControlConcurrency ControlProblem definitionConcurrency control problemsConcurrency control tools

Concurrency Control ProblemConcurrency Control ProblemObjective:

– Maximize work performed– Throughput: number of transactions processed

per unit timeConstraint:

– No interference: serial effect– Interference occurs on commonly manipulated

data known as hot spots

Lost Update ProblemLost Update Problem

Transaction A Time Transaction B Read SR (10) T1 T2 Read SR (10) If SR > 0 then SR = SR -1

T3

T4 If SR > 0 then SR = SR -1

Write SR (9) T5 T6 Write SR (9)

Uncommitted Dependency Uncommitted Dependency ProblemProblem

Transaction A Time Transaction B Read SR (10) T1 SR = SR - 1 T2 Write SR (9) T3 T4 Read SR (9) ROLLBACK T5

Inconsistent Retrieval ProblemsInconsistent Retrieval ProblemsInterference causes inconsistency among

multiple retrievals of a subset of dataIncorrect summary Phantom readNon repeatable read

Incorrect Summary ProblemIncorrect Summary ProblemTransaction A Time Transaction B Read SR1 (10) T1 SR1 = SR1 - 1 T2 Write SR1 (9) T3 T4 Read SR1 (9) T5 Sum = Sum + SR1 T6 Read SR2 (5) T7 Sum = Sum + SR2 Read SR2 (5) T8 SR2 = SR2 - 1 T9 Write SR2 (4) T10

Locking FundamentalsLocking FundamentalsFundamental tool of concurrency controlObtain lock before accessing an itemWait if a conflicting lock is held

– Shared lock: conflicts with exclusive locks– Exclusive lock: conflicts with all other kinds

of locksConcurrency control manager maintains

the lock table

Locking GranularityLocking GranularityDatabase

Table

Page

Row

Column

Index

Deadlock (Mutual Waiting)Deadlock (Mutual Waiting)

Transaction A Time Transaction B

XLock SR1 T1

T2 XLock SR2

XLock SR2 (wait) T3

T4 XLock SR1 (wait)

Deadlock ResolutionDeadlock ResolutionDetection

– Can involve significant overhead– Not widely used

Timeout– Waiting limit– Can abort transactions that are not deadlocked– Widely used although timeout interval is

difficult to determine

Two Phase Locking (2PL)Two Phase Locking (2PL)Protocol to prevent lost update problemsAll transactions must followConditions

– Obtain lock before accessing item– Wait if a conflicting lock is held– Cannot obtain new locks after releasing locks

2PL Implementation2PL Implementation

Time

Growing phase

BOT EOT

Shrinkingphase

Lock

s he

ld

Optimistic ApproachesOptimistic ApproachesAssumes conflicts are rareNo locksCheck for conflicts

– After each read and write– At end of transaction

Evaluation– Less overhead– More variability

Recovery ManagementRecovery ManagementDevice characteristics and failure typesRecovery toolsRecovery processes

Storage Device BasicsStorage Device BasicsVolatile: loses state after a shutdownNonvolatile: retains state after a shutdownNonvolatile is more reliable than volatile

but failures can cause loss of dataUse multiple levels and redundant levels

of nonvolatile storage for valuable data

Failure TypesFailure TypesLocal

– Detected and abnormal termination– Limited to a single transaction

Operating System– Affects all active transactions– Less common than local failures

Device– Affects all active and past transactions– Least common

Transaction LogTransaction LogHistory of database changesLarge storage overheadOperations

– Undo: revert to previous state– Redo: reestablish a new state

Fundamental tool of recovery management

Transaction Log ExampleTransaction Log Example

LSN TransNo Action Time Table Row Column Old New

1 101001 START 10:29 2 101001 UPDATE 10:30 Acct 10001 AcctBal 100 200 3 101001 UPDATE 10:30 Acct 15147 AcctBal 500 400 4 101001 INSERT 10:32 Hist 25045 * <1002,

500, …>

5 101001 COMMIT 10:33

CheckpointsCheckpointsReduces restart work but adds overhead

– Checkpoint log record– Write log buffers and database buffers

Checkpoint interval: time between checkpoints

Types of checkpoints– Cache consistent– Fuzzy

Other Recovery ToolsOther Recovery ToolsForce writing

– Checkpoint time– End of transaction

Database backup– Complete– Incremental

Recovery from a Media FailureRecovery from a Media FailureRestore database from the most recent

backupRedo all committed transactions since the

most recent backupRestart active transactions

Recovery TimelineRecovery Timeline

Time

T1

T5

T4

T3

T2

Checkpoint Failure

Recovery ProcessesRecovery ProcessesDepend on timing of database writesImmediate update approach:

– Before commit– Log records written first (write-ahead log

protocol)Deferred update approach

– After commit– Undo operations not needed

Immediate Update RecoveryImmediate Update RecoveryClass Description Restart Work T1 Finished before CP None T2 Started before CP;

finished before failure Redo forward from checkpoint

T3 Started after CP; finished before failure

Redo forward from checkpoint

T4 Started before CP; not yet finished

Undo backwards from most recent log record

T5 Started after CP; not yet finished

Undo backwards from most recent log record

Deferred Update RecoveryDeferred Update RecoveryClass Description Restart Work T1 Finished before CP None T2 Started before CP;

finished before failure Redo forward from first log record

T3 Started after CP; finished before failure

Redo forward from first log record

T4 Started before CP; not yet finished

None

T5 Started after CP; not yet finished

None

Transaction Design IssuesTransaction Design IssuesTransaction boundaryIsolation levelsDeferred constraint checkingSavepoints

Transaction Boundary Transaction Boundary DecisionsDecisionsDivision of work into transactionsObjective: minimize transaction durationConstraint: enforcement of important

integrity constraintsTransaction boundary decision can affect

hot spots

Registration Form ExampleRegistration Form Example

Transaction Boundary ChoicesTransaction Boundary ChoicesOne transaction for the entire formOne transaction for the main form and one

transaction for all subform recordsOne transaction for the main form and

separate transactions for each subform record

Isolation LevelsIsolation LevelsDegree to which a transaction is separated

from the actions of other transactionsBalance concurrency control overhead

with interference problemsSome transactions can tolerate

uncommitted dependency and inconsistent retrieval problems

Specify using the SET TRANSACTION statement

SQL Isolation LevelsSQL Isolation LevelsLevel XLocks SLocks PLocks Interference

Read uncommitted

None None None Uncommitted dependency

Read committed

Long Short None All except uncommitted dependency

Repeatable read

Long Long Short (S), Long (X)

Phantom reads

Serializable Long Long Long None

Scholar’s Lost UpdateScholar’s Lost UpdateTransaction A Time Transaction BObtain S lock on SR T1

Read SR (10) T2

Release S lock on SR T3

If SR > 0 then SR = SR -1 T4

T5 Obtain S lock on SR

T6 Read SR (10)

T7 Release S lock on SR

T8 If SR > 0 then SR = SR -1 Obtain X lock on SR T9

Write SR (9) T10

Commit T11

T12 Obtain X lock on SR

T13 Write SR (9)

Integrity Constraint TimingIntegrity Constraint TimingMost constraints checked immediatelyCan defer constraint checking to EOTSQL SET CONSTRAINTS statement

Save PointsSave PointsSome transactions have tentative actionsSAVEPOINT statement determines

intermediate pointsROLLBACK to specified save points

Workflow ManagementWorkflow ManagementWorkflow descriptionEnabling technologiesAdvanced transaction management

WorkflowsWorkflowsSet of tasks to accomplish a business

processHuman-oriented vs. computer-oriented

– Amount of judgment– Amount of automation

Task structure vs. task complexity– Relationships among tasks– Difficulty of performing individual tasks

Enabling TechnologiesEnabling TechnologiesDistributed object management

– Many kinds of non traditional data– Data often dispersed in location

Workflow modeling– Specification– Simulation– Optimization

Advanced Transaction Advanced Transaction ManagementManagementConversational transactionsTransactions with complex structureTransactions involving legacy systemsCompensating transactionsMore flexible transaction processing

SummarySummaryTransaction: user-defined collection of workDBMSs support ACID propertiesKnowledge of concurrency control and

recovery important for managing databasesTransaction design issues are importantTransaction processing is an important part

of workflow management