Research On The Integrated Testing Method Of DBMS Locking Mechanism

Database manage system (DBMS) plays key role in most huge application systems. The correctness and reliability of DBMS is the most important factor for those application systems running stably. Locking mechanism is an important part of DBMS, which supports the DBMS to process transactions concurrently, whose correctness and reliability would affect DBMS deeply. Thus, it is necessary to study integrated testing method of locking mechanism.The modules of locking mechanism couple other modules tightly. There are so many function-callings between them that it is hard to depart them. More ever, locking table, the key data structure of locking mechanism is very complex, there is concurrently processing in locking mechanism and the state of locking table is keeping changing when whole system is running. All of those things make it a hard work to hold a integrated testing on locking mechanism.A new integrated testing of locking mechanism is presented in this paper, which is based on the analysis of source code of a DBMS. The detail of the data structure, function list and the calling relationship with other modules are obtained. According to those information, the module of locking mechanism is departed from whole system. The stable fuctions are programmed to simulate some functions, which are called by functions in locking mechanism. Thus, the integrated testing won't be affected by the faults outside locking mechanism.Three kinds of test methods are applied to generate testing sequences, which include random testing method, prefix-based testing method and state machine based testing method. When random testing method is used, the"smoking testing"method should be used to verify the testing results. The prefix-based testing method generates a function calling sequence, which would make locking table transform to a certain state. According to the state, the sequences are generated based on the sensitive operators. Those sequences are executed and the states of locking table is monitored to aware whether the testing results are correct. Besides, the finite state machines can be drawn from the properties of locking mechanism, based on which the test cases can be generated. A semi-automatic testing framework is implemented, which uses all above methods. The experiment is hold and the result shows all those methods really work.