Dağıtılmış veritabanı sistemlerinde eşzamanlılık ve eşzamanlılık kontrolü

Abstract

SGT based pessimistic concurrency control. A SSG, that represents the conflicting operations from among those operations accepted to be scheduled, is maintained. The SSG mainted is ensured to be acyclic such that the execution is SR. There are basic SGT, conservative SGT, stict SGT 2. Optimistic concurrency control algorithms. They have the view that transactions, conflicts occur rare for most concurrent transaction execution. Therefore, no transaction is delayed or aborted during its execution as in optimisstic methods. Instead, each transaction is first executed to its commit point and then a validation test is performed to see wether or not there have been any data access conflicts that may have caused inconsistency of the database. The following methods are explained in detail: 1) 2PL based optimistic concurrency control, 2) TO based optimistic concurrency control, 3) SGT based optimistic concurrency control. 3) Hyprid method, it takes the combination of several mechanisms (i.e 2FL, TO and SGT) to do the concurrency control. The combination of 2PL and TO is explained in detail. At the end of this research on concurrency control in distributed database systems a central transaction manager (TM) is chosen and strict 2PL is used for currency control with timstamp ordering based Wait-Die rule to solve deadlocks. A program is made with this method and in the used program five distributed sites are assumed. In every site there exists five users. The data is assumed to be distributed and nonreplicated. The program is listed in Ek-II. Ek-I (fig.2.2.1) shows the flow and interconnection between the program's procedures. Program input, output and the flow of the program is explained in Bulgular section. The program is run on five different input files. Input and output files are listed in Ek-IH. In Tartışma and Sonuç section, the alternative mehods for concurrency control in distributed database systems are reevaluated. At the end of this reevaluation the run of he program, which is built based on 2PL and TO, is shown, see Ek.III for input and output files. -78-V. SUMMARY ` Concurrency and Concurrency Control In distributed Database Systems` The main problems in DDBMS are dassified in three principal groups: Concurrency Control Query Processing Reliability The fundamental problem facing the designers of DDBMSs is that of the correct control of concurrent accesses to distributed DBS by many different users. This can be viewed as the design of an approrpriate concurrency control algorithm. A solution of this problem influences the solution of the two remaining issues. Thus concurrency control algorithms are studied in this thesis. They are classified into three main parts: 1. Pessimistic concurrency control algorithms. They have the view that many transactions conflict with each other. Thus they synchronize access to shared data by concurrent transactions early in their execution life cycle. The following methods are explained in details: 2PL based pessimistic concurrency control. Each data item has a lock associated with it. The algorithms ensure that only one transaction can hdd a conflicting lock at a time, so only one transaction can access the item at a time. There are basic 2PL, strict 2FL, conservative 2PL, primary copy 2PL, primary site 2PL (Centralized) and decentralized 2PL (Distributed 2PL). 2PL methods result in deadlocks when implemented, thus deadlock management for 2PL is studied. There are deadlock prevention, deadlock avoidance based on timestamps, deadlock detection and resoultion and, centralized (Global) deadlock detection TO based pessimistic concurrency control. Each transaction is assigned a timestamp by the TM. The conflicting operations of transactions are ordered according to their timestamps. There are basic TO, basic TO with TWR, conservative TO and, strict TO. -77-SGT based pessimistic concurrency control. A SSG, that represents the conflicting operations from among those operations accepted to be scheduled, is maintained. The SSG mainted is ensured to be acyclic such that the execution is SR. There are basic SGT, conservative SGT, stict SGT 2. Optimistic concurrency control algorithms. They have the view that transactions, conflicts occur rare for most concurrent transaction execution. Therefore, no transaction is delayed or aborted during its execution as in optimisstic methods. Instead, each transaction is first executed to its commit point and then a validation test is performed to see wether or not there have been any data access conflicts that may have caused inconsistency of the database. The following methods are explained in detail: 1) 2PL based optimistic concurrency control, 2) TO based optimistic concurrency control, 3) SGT based optimistic concurrency control. 3) Hyprid method, it takes the combination of several mechanisms (i.e 2FL, TO and SGT) to do the concurrency control. The combination of 2PL and TO is explained in detail. At the end of this research on concurrency control in distributed database systems a central transaction manager (TM) is chosen and strict 2PL is used for currency control with timstamp ordering based Wait-Die rule to solve deadlocks. A program is made with this method and in the used program five distributed sites are assumed. In every site there exists five users. The data is assumed to be distributed and nonreplicated. The program is listed in Ek-II. Ek-I (fig.2.2.1) shows the flow and interconnection between the program's procedures. Program input, output and the flow of the program is explained in Bulgular section. The program is run on five different input files. Input and output files are listed in Ek-IH. In Tartışma and Sonuç section, the alternative mehods for concurrency control in distributed database systems are reevaluated. At the end of this reevaluation the run of he program, which is built based on 2PL and TO, is shown, see Ek.III for input and output files. -78-

V. SUMMARY ` Concurrency and Concurrency Control In distributed Database Systems` The main problems in DDBMS are dassified in three principal groups: Concurrency Control Query Processing Reliability The fundamental problem facing the designers of DDBMSs is that of the correct control of concurrent accesses to distributed DBS by many different users. This can be viewed as the design of an approrpriate concurrency control algorithm. A solution of this problem influences the solution of the two remaining issues. Thus concurrency control algorithms are studied in this thesis. They are classified into three main parts: 1. Pessimistic concurrency control algorithms. They have the view that many transactions conflict with each other. Thus they synchronize access to shared data by concurrent transactions early in their execution life cycle. The following methods are explained in details: 2PL based pessimistic concurrency control. Each data item has a lock associated with it. The algorithms ensure that only one transaction can hdd a conflicting lock at a time, so only one transaction can access the item at a time. There are basic 2PL, strict 2FL, conservative 2PL, primary copy 2PL, primary site 2PL (Centralized) and decentralized 2PL (Distributed 2PL). 2PL methods result in deadlocks when implemented, thus deadlock management for 2PL is studied. There are deadlock prevention, deadlock avoidance based on timestamps, deadlock detection and resoultion and, centralized (Global) deadlock detection TO based pessimistic concurrency control. Each transaction is assigned a timestamp by the TM. The conflicting operations of transactions are ordered according to their timestamps. There are basic TO, basic TO with TWR, conservative TO and, strict TO. -77-SGT based pessimistic concurrency control. A SSG, that represents the conflicting operations from among those operations accepted to be scheduled, is maintained. The SSG mainted is ensured to be acyclic such that the execution is SR. There are basic SGT, conservative SGT, stict SGT 2. Optimistic concurrency control algorithms. They have the view that transactions, conflicts occur rare for most concurrent transaction execution. Therefore, no transaction is delayed or aborted during its execution as in optimisstic methods. Instead, each transaction is first executed to its commit point and then a validation test is performed to see wether or not there have been any data access conflicts that may have caused inconsistency of the database. The following methods are explained in detail: 1) 2PL based optimistic concurrency control, 2) TO based optimistic concurrency control, 3) SGT based optimistic concurrency control. 3) Hyprid method, it takes the combination of several mechanisms (i.e 2FL, TO and SGT) to do the concurrency control. The combination of 2PL and TO is explained in detail. At the end of this research on concurrency control in distributed database systems a central transaction manager (TM) is chosen and strict 2PL is used for currency control with timstamp ordering based Wait-Die rule to solve deadlocks. A program is made with this method and in the used program five distributed sites are assumed. In every site there exists five users. The data is assumed to be distributed and nonreplicated. The program is listed in Ek-II. Ek-I (fig.2.2.1) shows the flow and interconnection between the program's procedures. Program input, output and the flow of the program is explained in Bulgular section. The program is run on five different input files. Input and output files are listed in Ek-IH. In Tartışma and Sonuç section, the alternative mehods for concurrency control in distributed database systems are reevaluated. At the end of this reevaluation the run of he program, which is built based on 2PL and TO, is shown, see Ek.III for input and output files. -78-