Database Management System Lecture - 44 © Virtual University of Pakistan

Uncommitted Update Problem TIME TA TB BAL t1 Read (BAL) ……. 1000 t2 BAL=BAL+100 …….. t3 Write (BAL) 1100 t4 t5 BAL=BAL*1.5 t6 2650 t7 ROLLBACK ? © Virtual University of Pakistan

Inconsistent Analysis One transaction operates many records, meanwhile another modifies some of them effecting result of first Suppose Tr-A computes interest on all accounts balances and other moves balance from one account to another © Virtual University of Pakistan

© Virtual University of Pakistan TIME TA TB t1 Read (BAL_A) (5000) ……. t2 INT = BAS_A * .05 …….. t3 TOT = TOT + INT t4 t5 Read (BAL_A) t6 BAL_A=BAL_A – 1000 t7 Write (BAL_A) t8 Read (BAL_E) (5000) t9 BAL_E = BAL_E + 1000 t10 Write (BAL_E) t11 Read (BAL_E) (6000) t12 INT = BAS_E * .05 t13 © Virtual University of Pakistan

© Virtual University of Pakistan Serial Execution An execution where transaction are executed in a serial Schedule or History For two transactions two serial executions or schedules are possible © Virtual University of Pakistan

© Virtual University of Pakistan Serial Execution TA, TB BAL TB, TA Read (BAL)A 50 Read (BAL)B (BAL = BAL – 10)A (BAL=BAL * 2)B Write (BAL)A 40 Write (BAL)B 100 80 90 © Virtual University of Pakistan

© Virtual University of Pakistan Serial Execution Different serial schedules may result in a different final DB state, BUT Serial execution is guaranteed to leave DB in a consistent state None of the three concurrent access problems © Virtual University of Pakistan

© Virtual University of Pakistan Lesson? Serial schedule is guaranteed correct, should we adopt it? NNNNNNo, that will be inefficient, under utilization of the resources Disliked by users © Virtual University of Pakistan

© Virtual University of Pakistan Interleaved Schedule We would prefer concurrent execution, multiple users accessing database at the same time An interleaved schedule consists of operations from different transactions © Virtual University of Pakistan

© Virtual University of Pakistan Schedule TA = {Read(X), Write(X), commit} TB = {Read(Y), Read(X), Write(Y), commit} S1 = {RA(X), WA(X), CA, RB(Y), RB(X), WB(Y), CB} S2 = {RB(Y), RB(X), WB(Y), CB, RA(X), WA(X), CA} S3 = {RB(Y), RB(X), RA(X), WB(Y), CB, WA(X), CA} S4 = {RA(X), RB(Y), RB(X), WA(X), CA, WB(Y), CB} © Virtual University of Pakistan

© Virtual University of Pakistan Interleaved Schedule Interleaved schedule provides concurrent execution BUT Three problems of concurrent access may be encountered © Virtual University of Pakistan

© Virtual University of Pakistan Real Problem Two transactions Writing different variables Reading same variable Accessing same variable and one of them writing Conflicting operations Conflicting transactions © Virtual University of Pakistan

© Virtual University of Pakistan Serializability An interleaved schedule is serializable if the final state of the schedule is equivalent to a serial schedule OR A schedule where conflicting operations are in a serial order © Virtual University of Pakistan

© Virtual University of Pakistan Serializability Two major approaches Locking Timestamping © Virtual University of Pakistan

© Virtual University of Pakistan Locking An object is locked before it performs any operation Two types of operations, two types of locks Shared and esclusive © Virtual University of Pakistan

© Virtual University of Pakistan Lock Compatibility TA holds TB requests Shared Exclusive YES NO © Virtual University of Pakistan

Database Management System Lecture - 44 © Virtual University of Pakistan