Simultaneous access of a shared record by multiple transactions leads to conflicts while writing. Such scenario generates the problems like lost update, dirty read, nonrepeatable read etc . In such a case, transaction need to be rolled back to get the system into a consistent state. To handle these conflicts Multiversion Concurrency control (MVCC) is used. MVCC has the ability to avoid the read-write conflicts by performing the read operations using the older version when the write request is in progress. When the multiple write operation concurrently executes then the transaction is aborted or rolled back. To address the problem of rollbacks, a methodology using the Precedence Graph generation based algorithm to reschedule the sequence of the transaction is proposed. In the proposed methodology whenever there is a system failure or network failure the transaction need not be started again, rather it is partially rolled back. The proposed methodology is executed and compared with the other MVCC approached on same set of transactions defined on shared and non shared data items. It is observed that the proposed methodology achieves better execution time as compared to the methods available in the literature
Multiversion Concurrency Control with the Precedence Graph Generation Algorithm
