Research Of Concurrency Control And Recovery In A Client/Server DBMS

Client/Server architecture is popular in database systems. The security protection is an important subject of DBMS research. It concerns database security, completeness, concurrency and recovery from failures. The latter two goals show the reason of research on concurrency control and recovery techniques.DBMS must guarantee ACID of transactions. Concurrency control mainly concerns consistency and isolation, while recovey techniques guarantee atomicity and durability of transactions. This paper focuses on concurrency control and recovery of page-server architechture of Client/Server DBMS.This paper classifies algorithms of maintaining client cache consisetency, and introduces some representative ones. The advantages and applicable range of these algorithms are discussed respectively, the shortcomings and problems existed are also analysed. Then the two major recovery techniques of Client/Server DBMS are discussed.On basis of studying page-server and varies of cache consistency algorithms, a P2P-based cache consistency algorithm called P2PBCC algorithm is presented. P2PBCC allows not only the server but also the clients to maintain a directory for each cached page. The server is partially removed from the critical path of concurrency control. P2PBCC reduces the load of server and block time of client transactions, increases throughput of overall system, and reduces the abort rate of transations. This paper also presents the normal processing of DBMS using P2PBCC, designs the recovery strategies including how to deal with transaction rollback, reatart recovey, process failure, client failure and server failure etc. Finally, we simulation some algorithms, and compare their character and performance. P2PBCC outperforms the other ones while there is data contention among clients.