Transaction Processing

Transcription

1 Transaction Processing As part of DB Implementation Techniques - Wintersemester 2010/2011 Prof. Dr.-Ing. Hagen Höpfner, Juniorprofessur Mobile Media, BU Weimar This set of slides is based on slides by G. Saake and A. Heuer.

2 Agenda } Transaction models } Transaction management } Recovery and Data safety } Distributed transactions Slide 2

3 Course Literature } Weikum and Vossen: Transactional Information Systems. Morgan Kaufmann Publishers (2002, ISBN: ) } Elmasri and Navathe: Fundamentals of Database Systems. Pearson Education, Inc. (4 th ed., 2004, ISBN ) Slide 3

4 Part 1: Transaction models } Transactions in multi-user systems } Transactional properties } Problems in multi-user systems } Serializability } Correctness in the presence of aborts Slide 4

5 1.1 Transactions in multi-user systems } Operational integrity } Multiple users access same data at the same time } Multiple application run simultaneously à Concurrent and competitively processes } Transactions as operands Slide 5

6 Examples } Parallel seat reservation by various travel agencies } It may be possible to sell a seat twice if two travel agencies consider it to be available. } Overwriting account operations of a bank } Statistical database operations } Updating database during computing statistics may result in wrong results. Slide 6

7 1.2 Transactional properties Slide 7

8 ACID conditions } Atomicity The sequence of operations either all take place or none do. } Consistency The database is transform from on valid state to another. } Isolation Each transaction believes to be the only transaction in the system. } Durability Once a transaction is committed it remains independent of failures. Slide 8

9 Commands of a transaction language } Begin-of-Transaction BOT } Successful end of a transaction commit } Abort of a transaction abort Slide 9

10 1.3 Problems in multi-user systems } Non-repeatable Read } Inconsistent Read } Dirty Read } The Phantom-Problem } Lost Update } Multi-user Anomaly } Cursor-Reference Problem Slide 10

11 Non-repeatable Read Example (1) } Covenant: X=A+B+C must hold at the end of transaction T 1 } X and Y are local variables } T i is transaction i } Constraint: A+B+C=0 Slide 11

12 Non-repeatable Read Example (2) T 1 T 2 X:=A; Y:=A/2; A:=Y; C:=C+Y; commit; X:=X+B; X:=X+C; commit; Slide 12

13 Inconsistent Read - Example T 1 T 2 read(x); X:=X-1; read(y); write(x); Y:=Y+1; read(x); read(y); write(y); Commit; Constraint: X+Y=0 Slide 13

14 Dirty Read - Example T 1 T 2 read(x); X:=X+100; write(x); read(x); Y:=Y+X; write(y); commit; abort; Slide 14

15 The Phantom-Problem - Example select count(*) into X from CUSTOMER; update CUSTOMER set Bonus=Bonus+10000/X; commit; T 1 T 2 insert into CUSTOMER values (6789, Lilo, Pause, ); commit; Slide 15

16 Lost Update - Example T 1 T 2 X read(x); 10 read(x); 10 X:=X+1; 10 X:=X+1; 10 write(x); 11 write(x); 11 Slide 16

17 Multi-user Anomaly Example (1) } Constraint: A=B } T 1 =á A:=A+10; B:=B+10;ñ } T 2 =á A:=A*1.1; B:=B*1.1;ñ Slide 17

18 Multi-user anomaly Example (2) T 1 T 2 A B read(a); A:=A+10; write(a); 20 read(a); A:=A*1.1; write(a); 22 read(b); B:=B*1.1; write(b); 11 read(b); B:=B+10; write(b); Slide 18

19 Cursor-Reference Problem Example (1) Moving cursor C to the next tuple that satisfies predicate P. (Tuple A) Fetch tuple T 1 T 2 Update tuple A regarding P. Slide 19

20 Cursor-Reference Problem Example (2) T 1 T 2 declare C cursor for select * from PRODUCTS where price<100; fetch C into update PRODUCTS set price=100 where ProdNo=42; Slide 20

21 Overview about problems } Updates disappear (lost update) } Reading invalid data (dirty read) } Non isolated transactions cause interferences (non repeatable read) Slide 21

22 1.4 Serializability } Introduction } Schedules } Kinds of serializablity } Comparison Slide 22

23 Introduction into serializability T 1 : read A; A:=A-10; write A; read B; B:=B+10; write B; T 2 : read B; B:=B-20; write B; read C; C:=C+20; write C; } Execution alternatives for two transactions: } Serial, e.g. T 1 before T 2 } Merged, e.g. alternating Slide 23

24 Merged execution examples Alternative 1 Alternative 2 Alternative 3 T 1 T 2 T 1 T 2 T 1 T 2 read A read A read A A:=A-10 read B A:=A-10 write A A:=A-10 read B read B B:=B-20 write A B:=B+10 write A B:=B-20 write B write B read B read B read B write B B:=B-20 read C B:=B+10 write B B:=B+10 read C read C C:=C+20 write B C:=C+20 write B C:=C+20 write C write C write C Slide 24

25 Results of the Alternatives A B C A+B+C Initial values Alternative Alternative Alternative Slide 25

26 Lock-Unlock Model } Ideas: } An object that should be modified has to be read first. } After the object was modified it is written. } Between unlock and lock no modification of the object is allowed. } T 1 : lock A; unlock A; lock B; unlock B; } T 2 : lock B; unlock B; lock C; unlock C; Slide 26

27 Lock-Unlock Model - Example Alternative 1 Alternative 2 Alternative 3 T 1 T 2 T 1 T 2 T 1 T 2 lock A lock A lock A unlock A lock B lock B lock B unlock A unlock A unlock B unlock B lock B lock B lock B unlock B unlock B lock C lock C lock C unlock B unlock B unlock C unlock C unlock C Slide 27

28 Lock-Unlock-Model and Function Semantics T 1 T 2 T 3 lock A lock B lock A lock B lock C lock C unlock A f 1 (A,B) unlock B f 3 (B,C) unlock C f 6 (A,C) unlock B f 2 (A,B) lock A unlock A f 7 (A,C) unlock C f 4 (A,B,C) unlock A f 5 (A,B,C) Slide 28

29 Merged Execution and Function Semantics - Example Step A B C (1) T 1 : lock A A 0 B 0 C 0 (2) T 2 : lock B A 0 B 0 C 0 (3) T 2 : lock C A 0 B 0 C 0 (4) T 2 : unlock B A 0 f 3 (B 0,C 0 )=σ 1 C 0 (5) T 1 : lock B A 0 σ 1 C 0 (6) T 1 : unlock A f 1 (A 0, σ 1 )=σ 2 σ 1 C 0 (7) T 2 : lock A σ 2 σ 1 C 0 (8) T 2 : unlock C σ 2 σ 1 f 4 (σ 2,B 0,C 0 )=σ 3 (9) T 2 : unlock A f 5 (σ 2,B 0,C 0 )=σ 4 σ 1 σ 3 (10) T 3 : lock A σ 4 σ 1 σ 3 (11) T 3 : lock C σ 4 σ 1 σ 3 (12) T 1 : unlock B σ 4 f 2 (A 0,σ 1 ) σ 3 (13) T 3 : unlock C σ 4 f 2 (A 0,σ 1 ) f 6 (σ 4,σ 3 ) (14) T 3 : unlock A f 7 (σ 4,σ 3 ) f 2 (A 0,σ 1 ) f 6 (σ 4,σ 3 ) Slide 29

30 Calculated values of the last example } A =f 7 (σ 4,σ 3 ) =f 7 (f 5 (f 1 (A 0,f 3 (b 0,C 0 )),B 0,C 0 ),f 4 (f 1 (A 0,f 3 (B 0,C 0 )),B 0,C 0 )) } B =f 2 (A 0,σ 3 )=f 2 (A 0,f 3 (B 0,C 0 )) } C =f 6 (σ 4, σ 3 ) =f 6 (f 5 (f 1 (A 0,f 3 (B 0,C 0 )),B 0,C 0 ),f 4 (f 1 (A 0,f 3 (B 0,C 0 )),B 0,C 0 )) Slide 30

31 Serializability A merged execution of multiple transactions is called serializable if and only if it has the same effects as the serial execution. Slide 31

32 Conflict Graph } unlock X lock X is called a conflict if there is no modification of X by another transaction in between. T 1 T 2 T 3 Slide 32

33 Schedules u 4 u 3 u 2 u 1 v 4 v 3 v 2 v 1 w 4 w 3 w 2 w 1 Scheduler w 2 v 3 w 1 u 1 v 2 v 1 Slide 33

34 Read/Write Model } A transaction T is a finite sequence of operations p i of the form r(x i ) or w(x i ): T=p 1 p 2 p 3 p n with p i {r(x i ),w(x i )} } A completed transaction ends in an abort a or a commit c: T=p 1 p n a or T=p 1...p n c Slide 34

35 Formal Semantics of Transactions } [[w(x)]]:=f(y 1, y n ) } [[T]]:={[[w(x i )]] w(x i ) is write op. in T} Slide 35

36 Formal Semantics - Explanation } Given: Transaction T reads a, b, and c and writes x } x depends on a, b, and c and T depends on x } E.g: T=r(a)r(b)r(c) x:=a+b+c w(x) à [[T]]={[[w(x)]]} with [[w(x)]]=f(a,b,c) Slide 36

37 Merged Execution } Let T be a the set of transactions T i. The set of all possible merged executions of all T i T is called SHUFFLE(T ). } All operations of a transaction T i are contained once per s SHUFFLE(T ). } Order in-between operations from the same transaction has been retained unchanged. Slide 37

38 SHUFFLE(T ) - Example } T 1 :=r 1 (x)w 1 (x) } T 2 :=r 2 (x)r 2 (y)w 2 (y) } T :={T 1, T 2 } } SHUFFLE(T )={ r 1 (x)w 1 (x)r 2 (x)r 2 (y)w 2 (y), r 2 (x)r 1 (x)w 1 (x)r 2 (y)w 2 (y), } Slide 38

39 Schedule } A schedule is a prefix of a complete schedule } Example: r 1 (x)r 2 (x)w 1 (x)r 2 (y)a 1 w 2 (y)c 2 schedule complete schedule Slide 39

40 Serial Schedule } A serial schedule s for T is a complete schedule of the form: s:=t ρ(1) T ρ(n) with index permutation ρ } Example: T 1 =r 1 (x)w 1 (x)c 1 and T 2 =r 2 (x)w 2 (x)c 2 } s 1 =T 1 T 2 =r 1 (x)w 1 (x)c 1 r 2 (x)w 2 (x)c 2 } s 2 =T 2 T 1 =r 2 (x)w 2 (x)c 2 r 1 (x)w 1 (x)c 1 Slide 40

41 Correctness Criterion } If the effects of the execution of a schedule s are equivalent to the effects of executing the serial schedule s containing the same operations (s v s), then s is called to be correct. } If a schedule s in equivalent to a serial schedule s, then s is serializable (to s ). Slide 41

42 View Serializability } A schedule is view-serializable if it is view-equivalent to a serial schedule. } Additional transactions required: } T 0 initially writes all involved objects } T finally reads all involved objects Slide 42

43 Read-from-Relation } Let be the relation in the schedule before, now r j (x) reads x from T i if and only if: } w i (x) r j (x) and } (w i (x) w k (x) w k (x) r j (x)) hold. } Read-from-Relation is defined as: } RF(s):={(T i,x,t j ) r j (x) reads x from T i )} Slide 43

44 View Equivalence } Two schedules s and s are view-equivalent if and only if: } op(s)=op(s ) Both schedules use the same operations including a and c. } RF(s)=RF(s ) Both schedules have the same Read-from-Relation. Slide 44

45 View Equivalence Example (1) } Given: two complete schedules s 1 and s 2 composed of two transaction T 1 and T 2 : } s 1 =r 1 (x)r 2 (y)w 1 (y)w 2 (y)c 1 c 2 } s 2 =r 1 (x)w 1 (y)r 2 (y)w 2 (y)c 2 c 1 } Adding T 0 and T : } s 1 =w 0 (x)w 0 (y)r 1 (x)r 2 (y)w 1 (y)w 2 (y)c 1 c 2 r (x) r (y) } s 2 =w 0 (x)w 0 (y)r 1 (x)w 1 (y)r 2 (y)w 2 (y)c 2 c 1 r (x) r (y) Slide 45

46 View Equivalence Example (2) } Generating the Read-from-Relations: } RF(s 1 )={(T 0,x,T 1 ),(T 0,y,T 2 ),(T 0,x,T ),(T 2,y,T )} } RF(s 2 )={(T 0,x,T 1 ),(T 1,y,T 2 ),(T 0,x,T ),(T 2,y,T )} } Obviously: RF(s 1 ) RF(s 2 ) holds } So s 1 and s 2 are not view-equivalent! Slide 46

47 View Serializability } Remember: A schedule is view-serializable if it is viewequivalent to a serial schedule. } VSR is the set of all view serializable schedules. Slide 47

48 View Serializability - Example } Possible serial schedules regarding s 1 are s and s : } s =T 1 T 2 =r 1 (x)w 1 (y)c 1 r 2 (y)w 2 (y)c 2 } RF(s )={(T 0,x,T 1 ),(T 1,y,T 2 ),(T 0,x,T ),(T 2,y, T )} } RF(s ) RF(s 1 ) } s =T 2 T 1 =r 2 (y)w 2 (y)c 2 r 1 (x)w 1 (y)c 1 } RF(s )={(T 0,y,T 2 ),(T 0,x,T 1 ),(T 0,x,T ),(T 1,y, T )} } RF(s ) RF(s 1 ) } So, s 1 is not view serializable. Slide 48

49 Conflict Serializability } Conflict relation C for s: } C(s):={(p,q) p conflicts with q and p q} } Conflict matrix: r i (x) w i (x) r j (x) - w j (x) - - Slide 49

50 Adjusted Conflict Relation } The adjusted conflict relation conf(s) does not represent aborted transactions: conf(s):=c(s)-{(p,q) (p t q t ) t aborted(s)} } Let aborted(s) be the set of aborted transactions in the schedule s. Slide 50

51 Conflict Relation - Example } Given: Schedule s: } s=r 1 (x)w 1 (x)r 2 (x)r 3 (y)w 2 (y)c 2 a 1 c 3 } Conflict Relation for s: } C(s)={(w 1 (x),r 2 (x)),(r 3 (y),w 2 (y))} } Adjusted conflict relation for s: } conf(s)={(r 3 (y),w 2 (y))} Slide 51

52 Conflict Equivalence } Two schedules s and s are called to be conflict equivalent if and only if: } op(s)=op(s ) and } conf(s)=conf(s ) hold. Slide 52

53 Conflict Equivalence Example (1) } Given: two complete schedules s and s : } s=r 1 (x)r 1 (y)w 2 (x)w 1 (y)r 2 (z)w 1 (x)w 2 (y) } s =r 1 (y)r 1 (x)w 1 (y)w 2 (x)w 1 (x)r 2 (z)w 2 (y) } Are s and s conflict equivalent? } Obviously op(s)=op(s ) holds. Slide 53

54 Conflict Equivalence Example (2) } Does conf(s)=conf(s ) hold, too? } conf(s)={ (r 1 (x),w 2 (x)), (w 2 (x),w 1 (x)), (r 1 (y),w 2 (y)), (w 1 (y),w 2 (y)) } } conf(s )={ (r 1 (x),w 2 (x)), (w 2 (x),w 1 (x)), (r 1 (y),w 2 (y)), (w 1 (y),w 2 (y)) } Slide 54

55 Conflict serializability } A schedule is conflict serializable if it is conflict equivalent to a serial schedule. } CSR is the set of all conflict serializable schedules. Slide 55

56 Conflict serializability Example } Given: } s=r 1 (x)w 1 (y)w 2 (x)w 1 (y)r 2 (z)w 1 (x)w 2 (y) } Serial schedule s 1? } conf(s)? } Is s conflict serializable? Slide 56

57 Conflict graph } Let s be a schedule. } A conflict graph G(s)=(V,E) for s is defined by: } V includes all transactions } E includes all edges between conflicting transactions: (t,t ) E t t ( p t) ( q t ) with (p,q) conf(s) Slide 57

58 Conflict graph Example T 1 T 2 T 3 r(y) r(u) r(y) w(y) w(x) w(x) w(z) w(x) T 1 T 2 T 3 s=r 1 (y)r 3 (u)r 2 (y)w 1 (y)w 1 (x)w 2 (x)w 2 (z)w 3 (x) Slide 58

59 Conflict Graph Properties } Serial schedules are represented by an acyclic graph. } For a given acyclic graph G(s) a serial schedule s can be found. So, s and s are conflict equivalent. (topological ordering) } A schedule s is not conflict serializable if G(s) is a cyclic graph. Slide 59

60 VSR vs. CSR } Holds CSR VSR or VSR CSR? } Given: } s=r 1 (y)r 3 (w)r 2 (y)w 1 (y)w 1 (x)w 2 (x)w 2 (z)w 3 (x)c 2 c 1 c 3 } Is s in VSR? } Is s in CSR? Proof: Weikum, Vossen (page 95) Slide 60

61 Problems at runtime } Typically we have to consider unfinished schedules. } Not committed transactions may abort. Slide 61

62 Closure Properties of Schedules (1) } } } } Let E be a class of schedules: E is prefix closed if for every schedule s in E each prefix of s is also in E. E is commit closed if for every schedule s in E CP(s) is also in E. E is prefix commit closed if it is prefix closed AND commit closed. Slide 62

63 Closure Properties of Schedules (2) } VSR-Schedules are not prefix commit closed! } CSR-Schedules are prefix commit closes! Slide 63

64 VSR vs. CSR Example (1) } Given: } s=w 1 (x)w 2 (x)w 2 (y)c 2 w 1 (y)c 1 w 3 (x)w 3 (y)c 3 } Is s in VSR and if so show that VSR is not prefix closed? } Is s in CSR and if so Is CSR prefix commit closed? Slide 64

65 So far summery } } } } } } Serial View equivalent } s and s contain the same operations } s and s have the same RF-Relation View serializable } s may be a serial schedule } s is view serializable if it is view equivalent to s Conflict equivalent } s and s contain the same operations } s and s have the same conflict relation Conflict serializable } s may be a serial schedule } s is conflict serializable if it is conflict equivalent to s Prefix commit closed } All possible prefixes of a committed transaktion have the same properties as the completed schedule. Slide 65

66 Exercise 1 } s 1 =r 1 (x)w 1 (x)r 2 (x)r 3 (z)w 3 (x)r 1 (z) } s 2 =w 1 (x)r 2 (x)w 2 (z)r 3 (z)w 3 (y)w 1 (y) } s 3 =r 1 (x)w 1 (x)w 2 (x)w 3 (z)r 3 (y)r 3 (x) } s 4 =w 2 (x)w 1 (x)w 1 (y)w 2 (y)w 1 (y)w 3 (z) } s 5 =w 2 (x)w 2 (y)r 1 (x)r 1 (y)w 1 (y)w 3 (z) } VSR? } CSR? Slide 66

67 1.5 Aborts and Fault tolerance } s=r 1 (x)w 1 (x) r 2 (x) a 1 w 2 (x)c 2 } s VSR and s CSR } BUT: from a fault tolerance view points is not acceptable Slide 67

68 Recoverable Schedules RC } s is recoverable if: (T i reads from T j in s) (c i s) (c j c i ) } s=r 1 (x)w 1 (x)r 2 (x)a 1 w 2 (x)c 2 Slide 68

69 Recoverable Schedule Example } s 1 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)r 2 (y)w 2 (y)c 2 w 1 (z)c 1 } T 2 reads y from T 1, but c 2 appears before c 1 à s1 RC } s 2 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)r 2 (y)w 2 (y)w 1 (z)c 1 c 2 } s 1 RC Slide 69

70 Is RC enough? } s 2 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)r 2 (y)w 2 (y)w 1 (z)c 1 c 2 } What happens if T 1 aborts: s 2 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)r 2 (y)w 2 (y)w 1 (z)a 1 c 2 } T 2 must also abort à cascading aborts Slide 70

71 Avoiding cascading aborts ACA } A schedule s avoids cascading aborts if (T i reads x from T j in s) (c j r i (x)) holds. } A transaction is only allowed to read from already committed transactions. Slide 71

72 Avoiding cascading aborts Example } s 2 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)r 2 (y)w 2 (y)w 1 (z)a 1 c 2 à s 2 ACA } s 3 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)w 1 (z)c 1 r 2 (y)w 2 (y)c 2 à s 3 ACA Slide 72

73 Before image problem in ACA DB state Operation x=1 (initial value) x=2 w 1 (x 2) [BF =1] x,t 1 x=3 w 2 (x 3) [BF =2] x,t 2 a 1 : recover w 1 (x 2) using BF x :=1 But w 2 must remain! x=3 w 2 (x 3) [BF =3] x,t 2 a 2 : recover w 2 (x 3) using BF x :=?? Slide 73

74 Strict schedules ST } A schedule s is strict if (w j (x) p i (x) j i) (a j p i (x) c j p i (x), (p i {r,w})) holds. } It is not allowed to read or write a written object from a not yet finished transaction. Slide 74

75 Strict schedules Example } s 3 =w 1 (x)w 1 (y)r 2 (u)w 2 (x)w 1 (z)c 1 r 2 (y)w 2 (y)c 2 à s 3 ST } s 4 =w 1 (x)w 1 (y)r 2 (u)w 1 (z)c 1 w 2 (x)r 2 (y)w 2 (y)c 2 à s 4 ST Slide 75

76 CSR, VSR, RC, ACA, ST, Serial RC ACA ST Acceptable reparable VCR CSR serial Strict locking Slide 76

77 Exercise 2 } s 1 =r 1 (x)w 1 (x)r 2 (x)c 1 c 2 } s 2 =r 1 (x)w 1 (x)r 2 (x)c 2 c 1 } s 3 =r 1 (x)w 1 (x)c 1 r 2 (x)c 2 } s 4 =r 1 (x)w 1 (x)w 2 (x)c 1 c 2 } s 5 =r 1 (x)w 1 (x)r 2 (y)c 1 c 2 } s 6 =r 2 (x)r 1 (x)w 1 (x)c 1 w 2 (x)c 2 } s 7 =r 2 (x)r 1 (x)w 1 (x)w 2 (x)c 1 c 2 } s 8 =w 2 (x)w 1 (x)w 1 (y)c 1 r 2 (y)c 2 } s 9 =w 2 (x)w 1 (x)w 1 (y)r 2 (y)c 1 c 2 } s 10 =w 2 (x)w 1 (x)w 1 (y)r 2 (y)c 2 c 1 } s 11 =r 1 (x)w 2 (x)r 2 (y)w 1 (y)c 1 c 2 } s 12 =w 2 (x)r 1 (x)w 1 (y)w 2 (y)c 1 c 2 } s 13 =w 2 (x)w 1 (x)w 1 (y)w 2 (y)c 1 c 2 } s 14 =w 2 (x)w 1 (x)w 1 (y)w 2 (y)r 1 (y)c 1 c 2 } CSR?, RC?, ACA?, ST?, serial? Slide 77

78 Part 2: Transaction Management } Architecture } Locking based approaches Slide 78

79 The Scheduler T 1 T 2 T n Transaction Manager (TM) Scheduler (SC) Correct Schedule (r, w, c, a) Storage Manager (SM) Recovery Manager (RM) Buffer Manager (BM) Slide 79

80 Operations } BOT (Begin of Transaction) } EOT (End of Transaction) } r (read) } w (write) } a (abort) } c (commit) t={r w}*(a c) Slide 80

81 States of a Transaction active execute initial BOT running restart retry reject aborted stopped stop delay delayed EOT committed Slide 81

82 Optimistic Scheduler } A scheduler is called to be optimistic, if it allows the appearance of conflicts but tries to recognise and to resolve them. } Maximal parallelisation of transactions } Risk: Conflicts my appear at the END of a transaction Slide 82

83 Pessimistic (conservative) Scheduler } A scheduler is called to be pessimistic or conservative if it tries to avoid conflicts but accepts delayed transactions. } Lower parallelisation rate } Minimize recovery costs } Extreme: no parallelisation Slide 83

84 2.2 Locking based approaches } Read and Write locks as follows: } rl(x): read lock regarding object x } wl(x): write lock for object x } Unlock operations ru(x) and wu(x) are often considered together as u(x). Slide 84

85 Lock discipline } Write access w i (x) only after wl i (x) possible } Read access r i (x) only after rl i (x) or wl i (x) possible } Locking of unlocked objects only } wl i (x) after rl i (x) possible } Locks of one kind must be unique } u i (x) by T i à new rl i (x) or wl i (x) by T i is prohibited } commit after all unlocks only Slide 85

86 Deadlock } Alternatives: wl(x) delay à wl(y) t 1 t 2 wl(y) wl(x) DEADLOCK! } Recognise and remove deadlocks } Avoid deadlocks ß delay Slide 86

87 Recognition and Elimination } Waits-for graph } Lock elimination à abort a transaction: } Number of resolved loops } Duration / Length of a transaction } Cost for recovering a transaction } Importance of a transaction, Slide 87

88 Livelock-Problem } T 1 locks A. } T 2 tries to lock A but has to wait. } T 3 tries to lock A but has to wait, too. } T 1 unlocks A. } T 3 gets the next slot and locks A. } T 2 tries to lock A but has to wait. } T 4 tries to lock A but has to wait, too. } T 3 unlocks A. } T 4 gets the next slot and locks A. } Slide 88

89 2.3 Locking protocols T 1 T 2 wl(x) w(x) u(x) wl(x) w(x) u(x) wl(y) w(y) u(y) wl(y) w(y) w(y) Slide 89

90 Two-Phase-Locking-Protocol #locks lock unlock time Slide 90

91 Ignoring 2PL may result in conflicts T 1 T 2 u(x) wl(x) wl(y) u(x) u(y) wl(y) Slide 91

92 Strict Two-Phase-Locking- Protocol #locks lock unlock S2PL Avoids cascading aborts time Slide 92

93 Conservative Two-Phase- Locking-Protocol #locks lock unlock C2PL Avoid deadlocks time Slide 93

94 Conservative and Strict Two- Phase-Locking-Protocol #locks time lock unlock CS2PL Slide 94

95 2.4 Lock Granulates } Granulate Hierarchy in Databases } Hierarchical Locks } Tree protocols for tree indexes Slide 95

96 Granulate Hierarchy in relational Databases Database Relation Tuple Attribute Database File Cluster Page Logical Physical Slide 96