Parallel and Distributed Transaction Processing

Transaction Concept

▪ A transaction is a unit of program execution that accesses and

possibly updates various data items.
▪ E.g., transaction to transfer $50 from account A to account B:
1. read(A)
2. A := A – 50
3. write(A)
4. read(B)
5. B := B + 50
6. write(B)
▪ Two main issues to deal with:
• Failures of various kinds, such as hardware failures and system
crashes
• Concurrent execution of multiple transactions

2
 Required Properties of a Transaction

▪ Transaction to transfer $50 from account A to account B:

1. read(A)
2. A := A – 50
3. write(A)
4. read(B)
5. B := B + 50
6. write(B)
▪ Atomicity requirement
• If the transaction fails after step 3 and before step 6, money will be
“lost” leading to an inconsistent database state
▪ Failure could be due to software or hardware
• The system should ensure that updates of a partially executed
transaction are not reflected in the database

3
 Required Properties of a Transaction

▪ Transaction to transfer $50 from account A to account B:

1. read(A)
2. A := A – 50
3. write(A)
4. read(B)
5. B := B + 50
6. write(B)

▪ Durability requirement — once the user has been notified that the
transaction has completed (i.e., the transfer of the $50 has taken place),
the updates to the database by the transaction must persist even if there
are software or hardware failures.

4
 Required Properties of a Transaction (Cont.)
▪ Transaction to transfer $50 from account A to account B:
1. read(A)
2. A := A – 50
3. write(A)
4. read(B)
5. B := B + 50
6. write(B)

▪ Consistency requirement in above example:

▪ The sum of A and B is unchanged by the execution of the transaction
▪ A transaction, when starting to execute, must see a consistent database.
▪ During transaction execution the database may be temporarily inconsistent.
▪ When the transaction completes successfully the database must be consistent

5
 Required Properties of a Transaction (Cont.)

▪ Isolation requirement — if between steps 3 and 4, another transaction T2 is allowed

to access the partially updated database, it will see an inconsistent database (the sum A
+ B will be less than it should be).

T1 T2
1. read(A)
2. A := A – 50
3. write(A)
read(A), read(B), print(A+B)
4. read(B)
5. B := B + 50
6. write(B
▪ Isolation can be ensured trivially by running transactions serially
• That is, one after the other.

6
 Required Properties of a Transaction (Cont.)

▪ Isolation requirement — if between steps 3 and 6, another transaction T2 is allowed

to access the partially updated database, it will see an inconsistent database (the sum A
+ B will be less than it should be).

T1 T2
1. read(A)
2. A := A – 50 However, executing
3. write(A) multiple transactions
read(A), read(B), print(A+B) concurrently has
significant benefits.
4. read(B)
5. B := B + 50
6. write(B
▪ Isolation can be ensured trivially by running transactions serially
• That is, one after the other.

7
 Transaction State

▪ Active – the initial state; the transaction stays in this state while it is
executing
▪ Partially committed – after the final statement has been executed.
▪ Failed -- after the discovery that normal execution can no longer proceed.
▪ Aborted – after the transaction has been rolled back and the database
restored to its state prior to the start of the transaction. Two options after it
has been aborted:
• Restart the transaction
▪ can be done only if no internal logical error
• Kill the transaction
▪ Committed – after successful completion.

8
Transaction State (Cont.)

9
 Distributed Transactions

▪ Local transactions
• Access/update data at only one database
▪ Global transactions
• Access/update data at more than one database
▪ Key issue: how to ensure ACID properties for transactions in a system with
global transactions spanning multiple database

10
 Distributed Transactions

▪ Transaction may access data at several sites.

• Each site has a local transaction manager
• Each site has a transaction coordinator
▪ Global transactions submitted to any transaction coordinator

11
 Distributed Transactions
Each transaction coordinator is responsible
for:
▪ Starting the execution of transactions
that originate at the site.
▪ Distributing subtransactions at
appropriate sites for execution.
▪ Coordinating the termination of each
Site
transaction that originates at the site 4TC4
• transaction must be committed at
all sites or aborted at all sites. Site TM Site
1TC1 4 3TC3

TM TM
1 3

Site
2TC2

TM
2

12
 Distributed Transactions

▪ Each local transaction manager

responsible for:
• Maintaining a log for recovery
purposes
• Coordinating the execution
and commit/abort of the
transactions executing Site
at that site. 4TC4

Site TM Site
Question 5-1 1TC1 4 3TC3
Site 3 has initiated Transaction T to transfer
Tk. 1000 from account P at site 4 to account TM TM
Q at site 2. Discuss the responsibilities of 1 3
concerned transaction coordinator and
transaction managers.
Site
2TC2

TM
2

13
 System Failure Modes

▪ Failures unique to distributed systems:

• Failure of a site.
• Loss of massages
▪ Handled by network transmission control protocols such as TCP-IP
• Failure of a communication link
▪ Handled by network protocols, by routing messages via alternative
links
• Network partition
▪ A network is said to be partitioned when it has been split into two or
more subsystems that lack any connection between them
• Note: a subsystem may consist of a single node
▪ Network partitioning and site failures are generally indistinguishable.

14
 Coordinator

Site 1 Site 2
Transaction: Transfer Tk. 5000 from account A to
TC 1 TC 2 account B. A in site 2 and B in site 3. Transaction
has been initiated from site 1.
TM1 1 TM 2

LOG LOG

<Start T> by TC1

Site 3 Lock (A) request by TC1 granted by TM2
READ (A)
TC 3
A = A- 5000
TM 3 Write (A)
Lock (B) request by TC1 granted by TM3
READ (B)
LOG B=B+5000
WRITE (B)
Question 5-2: Why is COMMIT COMMIT (T)
protocol needed to COMMIT the UNLOCK (A)
transaction T initiated by TC1? Explain. UNLOCK (B)

15
 Commit Protocols

▪ The two-phase commit (2PC) protocol is widely used

▪ Three-phase commit (3PC) protocol avoids some drawbacks of 2PC, but is

more complex

▪ Consensus protocols solve a more general problem, but can be used for
atomic commit

▪ The protocols we study all assume fail-stop model – failed sites simply stop
working, and do not cause any other harm, such as sending incorrect
messages to other sites.

16
 Two Phase Commit Protocol (2PC)

▪ Execution of the protocol is initiated by the coordinator after the last step of
the transaction has been reached.

▪ The protocol involves all the local sites at which the transaction executed

▪ Protocol has two phases

▪ Let T be a transaction initiated at site S1 and let the transaction coordinator at

S1 be TC1

17
Coordinator <Ready T>
Site 2
Phase 1: Obtaining a Decision
Site 1

TC 1 TC 2 Coordinator asks all participants to prepare to

<Prepare
T> commit transaction T
TM1 1 TM1 2

TC1 adds the records <prepare T> to the log

<Prepar <ready
and forces log to stable storage
e T> T>
sends prepare T messages to all sites at which T
<Prepare
executed
T> Site 3 <Ready T>

TC 3 Upon receiving message, transaction manager at site

determines if it can commit the transaction or not
TM1 3
Decision1: All sites wants to COMMIT
• add the record <ready T> to the log
<ready • force all records for T to stable storage
T> • send ready T message to TC1

18
Coordinator <Ready T>
Site 2
Site 1 Phase 2: Recording the Decision
TC 1 TC 2
<Prepare Transaction COMMIT by TC1
T> TM1 2
TM1 1
• T can be committed of TC1 received a
<commit
T>
<ready T> message from all the
<Prepar <ready participating sites.
e T> T> • Coordinator adds a decision record,
<comm
<Prepare <commit T> to the log and forces record
it T>
T> Site 3 <Ready T> onto stable storage.
• Once the record stable storage it is
<commit TC 3
T> irrevocable (even if failures occur)
TM1 3 • Coordinator sends a message to commit
to each participant informing it of the
decision.
<ready • Participants take appropriate action
T>
locally.

Question 5-3: Write down the activities of the coordinator in phase 1 and phase 2 for the case
when all the sites including coordinator wants to commit

19
Coordinator <Ready T>
Site 2
Site 1 Phase 2: Recording the Decision
TC 1 TC 2 Transaction ABORT by TC1
<Prepare
T> TM1 2
TM1 1 • Even after receiving <ready T> messages
<abort T> by TC1 from all the participating sites, TC1
<Prepar has the authority to abort the transaction T.
<ready
e T> • Coordinator adds a decision record, <abort
T>
<abort T> to the log and forces record onto stable
T> <Prepare
T> Site 3 <Ready T>
storage.
• Once the record stable storage it is
<abort T> TC 3 irrevocable.
• Coordinator sends a message to abort to
TM1 3
each participant informing it of the decision.
• Participants take appropriate action locally.
<ready
T>

20
Coordinator <Ready T>
Site 2
Phase 1: Obtaining a Decision
Site 1

TC 1 TC 2 Coordinator asks all participants to prepare to

<Prepare
T> commit transaction T
TM1 1 TM1 2

TC1 adds the records <prepare T> to the log

<Prepar <ready
and forces log to stable storage
e T> T>
sends prepare T messages to all sites at which T
<Prepare
executed
T> Site 3 <abort T>

TC 3 Upon receiving message, transaction manager at site

determines if it can commit the transaction or not
TM1 3
Decision2: One site don’t want to COMMIT (Site3)
• add a record <no T> to the log and
<no T> • send abort T message to TC1

21
Coordinator <Ready T>
Site 2
Site 1 Phase 2: Recording the Decision
TC 1 TC 2
<Prepare Transaction must be aborted by TC1
T> TM1 2
TM1 1
• T must be aborted by TC1 if it received a
<abort T>
<no T> message from any of the
<Prepar <ready participating sites.
e T> T> • Coordinator adds a decision record,
<abort
<Prepare <abortT> to the log and forces record
T>
T> Site 3 <abort T> onto stable storage.
• Once the record stable storage it is
<abort T> TC 3
irrevocable (even if failures occur)
TM1 3 • Coordinator sends a message to abort to
each participant informing it of the
decision.
<No • Participants take appropriate action
T>
locally.

Discussion 5-4: Write down the activities of the site 3 in phase 1 and coordinator in phase 2 in
the case when site 3 do not want to commit.

22
Two-Phase Commit Protocol

23
 Handling of Failures - Site Failure

Site 2 was failed

When site S2 recovers, it examines its log to
determine the fate of
Site
transactions active at the time of the failure.
4TC4

Case 1: Log contain <commit T> record: Site TM Site

site 2 executes redo (T) 1TC1 4 3TC3

Case 2: Log contains <abort T> record: TM TM

1 3
site 2 executes undo (T)

Case 3: Log contains <ready T> record: Site

site must consult TC1 to determine the fate of T. 2TC2
If T committed, redo (T)
TM
If T aborted, undo (T) 2

24
 Handling of Failures - Site Failure

Site
Site 2 was failed 4TC4
Case 4: The log contains no control records
concerning T implies that S2 failed before Site TM Site
responding to the prepare T message from TC1. 1TC1 4 3TC3

TM TM
Since the failure of S2 precludes the sending of such 1 3
a response TC1 must abort T

S2 must execute undo (T) Site

2TC2

TM
2

25
 Handling of Failures - Coordinator Failure
If coordinator fails while the commit protocol for
T is executing then participating sites must
decide on T’s fate
Site
4TC4
Case 1: If an active site contains a <commit T>
record in its log, then T must be committed. Site TM Site
1TC1 4 3TC3
Case 2: If an active site contains an <abort T>
TM TM
record in its log, then T must be aborted. 1 3

Case 3: If some active participating site does not

Site
contain a <ready T> record in its log, then the failed
2TC2
coordinator Ci cannot have decided to commit T.
Can therefore abort T. TM
2

26
 Handling of Failures - Coordinator Failure
If coordinator fails while the commit protocol for
T is executing then participating sites must
decide on T’s fate
Site
4TC4
Case 4: If none of the above cases holds, then all
active sites must have a <ready T> record in their Site TM Site
logs, but no additional control records (such as 1TC1 4 3TC3
<abort T> or <commit T>). In this case active sites
must wait for Ci to recover, to find decision. TM TM
1 3

Blocking problem: active sites may have to wait Site

for failed coordinator to recover. 2TC2

TM
Question 5-5: Write a discussion on coordinator failure 2
cases and blocking problem

27
 Recovery and Concurrency Control

In-doubt transactions: When a failed site is recovered, the transactions those have a
<ready T>, but neither a <commit T>, nor an <abort T> log record.

The recovering site must determine the commit / abort status of In-doubt transactions by
contacting other sites;

1. this can slow the recovery (How?)

o Because after recovery, all active transactions must have redo or undo based on
Commit / abort status

2. Or can block the recovery (How?)

o In case of coordinator failure, site cannot decide commit/abort and indefinite wait
for coordinator

28
 Recovery and Concurrency Control

▪ For faster recovery, Recovery algorithms can note lock information in the
log.
• Instead of <ready T>, write out <ready T, L> L = list of locks held by T
when the log is written (read locks can be omitted).
▪ A Transaction T5 has Read lock (A), Write Lock (D) and Write Lock
( E) at site 3. T5 has Write Lock ( G) at site 2. T5 is initiated by site 1.
The data item A, D and E are in site 3. Site 3 wants to commit T5.
What log record will be written by site 3? How will it be used for
recovery?
▪ <ready T5, {D,E}>
• For every in-doubt transaction T, all the locks noted in the
<ready T, L> log record are reacquired.
▪ After lock reacquisition, transaction processing can resume; the commit or
rollback of in-doubt transactions is performed concurrently with the execution
of new transactions.

29
 Avoiding Blocking Using Consensus

▪ Blocking problem of 2PC is a serious concern

▪ How it arises?
Site
▪ T1 is initiated by site 1 (coordinator)
4TC4
▪ T1 has data from site 2, 3 and 4. Site 2, 3 and
4 are ready to commit and has written <ready Site TM Site
T1> to their log. 1TC1 4 3TC3

TM TM
1 3
Just after that, site 1 failed.
▪ Site 2, 3 and 4 are blocked for T1 and wait
until site 1 is restored. Site
2TC2
▪ Idea: involve multiple nodes in decision
process, so failure of a few nodes does not
TM
cause blocking as long as majority don’t fail 2

30
 Avoiding Blocking Using Consensus

Solution of Blocking Problem

▪ Distributed consensus protocol
Site
4TC4
▪ A set of n nodes need to agree on a decision
▪ Inputs to make the decision are provided to Site TM Site
1TC1 4 3TC3
all the nodes, and then each node votes on the
decision
TM TM
▪ The decision should be made in such a way 1 3
that all nodes will “learn” the same value for
the transaction even if some nodes fail during Site
the execution of the protocol. 2TC2

TM
2

31
 Using Consensus to Avoid Blocking

Solution of Blocking Problem

▪ Distributed consensus protocol
Site
4TC4
▪ After getting response from 2PC participants,
coordinator can initiate distributed consensus Site TM Site
protocol by sending its decision to a set of 1TC1 4 3TC3
participants who then use consensus protocol
to commit the decision TM TM
1 3
▪ If coordinator fails before informing all
consensus participants
Site
• Choose a new coordinator, which follows
2TC2
2PC protocol for failed coordinator
• If a commit/abort decision was made as TM
long as a majority of consensus 2
participants are accessible, decision can be
found without blocking

32
 Using Consensus to Avoid Blocking

Solution of Blocking Problem

▪ Distributed consensus protocol
Site
▪ Case 1: Coordinator (Site 1) failed before 4TC4
sending commit proposal to site 2, 3 and 4.
Site TM Site
1TC1 4 3TC3
Steps:
1. A new coordinator will be selected (e.g., site TM TM
1 3
3).
2. Site 3 will initiate 2PC on behalf of site 1
3. Commit/abort will be performed as per Site
2TC2
consensus protocol.
TM
No blocking of site 2, 3 and 4 for T1. 2

33
 Using Consensus to Avoid Blocking

Solution of Blocking Problem

▪ Distributed consensus protocol
Site
▪ Case 2: Coordinator (Site 1) failed after 4TC4
sending commit proposal to site 2, 3 and 4.
Site TM Site
1TC1 4 3TC3

▪ Commit/abort will be performed as per TM TM

1 3
consensus protocol.

No blocking of site 2, 3 and 4 for T1. Site

2TC2

TM
2

34
 Distributed Transactions via Persistent Messaging

▪ Notion of a single transaction spanning

multiple sites is inappropriate for many Persistent messaging systems are
applications systems that provide transactional
properties to messages
• E.g., transaction crossing an
organizational boundary
Persistent messages are guaranteed to
• Latency of waiting for commit from be delivered exactly once
remote site

▪ Problem to solve 4-2: There is an account

Debit Tk. 5000 from account A and send a
A under organization 1 and an account B
message to organization 2
under organization 2. Transfer Tk. 5000
from account A to account B using Organization 2 receives message and credits
persistent messaging protocol. Tk. 5000 to account B

35
 Distributed DBMS Environment
Organization 1

Site 1
Site 2

Site 5
Communication
Network

Organization 2

Site 4 Site 3

36
 Persistent Messaging

Atomicity issue
Once transaction sending a message is
committed, message must guaranteed to
be delivered

Guarantee as long as destination site is up

and reachable, code to handle
undeliverable messages must also be
available
e.g., credit money back to source
account.

If sending transaction aborts, message

must not be sent

37
 Error Conditions with Persistent Messaging
▪ Code to handle messages has to take care of variety of failure situations (even
assuming guaranteed message delivery)

▪ E.g., if destination account does not exist, failure message must be sent back to source
site

▪ When failure message is received from destination site, or destination site itself does
not exist, money must be deposited back in source account
▪ Problem if source account has been closed
• get humans to take care of problem

38
 Persistent Messaging Implementation

Problem to solve 4-2: There is an account A under organization 1 and an account B under
organization 2. Transfer Tk. 5000 from account A to account B using persistent messaging
protocol.

39
 Persistent Messaging (Cont.)

▪ Receiving site may get duplicate messages after a very long delay

▪ To avoid keeping processed messages indefinitely

• Messages are given a timestamp

• Received messages older than some cutoff are ignored

• Stored messages older than the cutoff can be deleted at receiving site

40
 Persistent Messaging (Cont.)

▪ Workflows provide a general model of transactional processing involving multiple

sites and possibly human processing of certain steps

▪ E.g., when a bank receives a loan application, it may need to

• Contact external credit-checking agencies

• Get approvals of one or more managers

and then respond to the loan application

▪ Persistent messaging forms the underlying infrastructure for workflows in a

distributed environment

41