Research On Dynamic Recovery Technology Based On Application Semantics

In recent years, intrusion tolerance, as a database security technology, has been gettingmore and more researchers’ attentions. Current intrusion tolerant databases focus on theresearch of tracking dependence between transactions, consistency and efficiency of transactionrecovery, etc. they did not consider transactions’ application environment, which may not allowdatabase to recover some transactions. How to keep the result of rigid transactions according tosemantic information, and try to recover other infected transactions, expecting the result ofrecovery can match the state of the world. It is significant.This paper puts forward the concept of the rigid transaction. It refs to infected transactionswhich can not be recovered. Rigid transactions are caused by its application environment.Transactions which are relied on by rigid transactions are called semi-rigid transactions.Recovery of semi-rigid transactions may lose the result of rigid transactions. In order todistinguish these two kinds of transactions from infected transactions, application semanticmechanism is introduced. The mechanism contains the design of semantic analysis algorithm,the design of semantic information log, designing the way interact with recovery module andso on.To deal with rigid transactions and semi-rigid transactions, we design transaction classifierin the dynamic recovery module. it is used to divide infected transaction set. We also redesignthe recovery algorithm. The algorithm eliminates the influence of malicious transactions asmuch as possible, on the base of reserving the rigid transactions.Finally, the system is realized in postgres. In the experiment part, we design experimentsaccording to the function and performance. Function experiment check whether the systemrecover different kinds of transactions in different ways according to the design, performanceexperiment check the performance influence to database when the system is in different state.