A transaction management scheme with mark in decision supporting environments

In a legacy operational system, a transaction processing system resolves the data conflicts between transactions by using locks or timestamps. As a result, one of the conflicting transactions should be blocked to resolve the data conflict between the transactions. Queries (or read-only transactions) are not the exceptions. In a query-intensive system such as a decision support system, many data items are read by one query and the length of each transaction is long as usual. If queries execute their operations without being interfered by update transactions, then the system can achieve a higher performance than the existing transaction processing systems. We propose a new transaction scheduling algorithm which eliminates the conflicts between queries and update transactions by using a new version selection scheme, called mark. The mark scheme allows that each transaction has its own version selection point. In the proposed protocol, new version selection rules and an efficient version maintenance scheme is also presented. The proposed protocol is based on a locking scheme and it provides a method to resolve deadlocks caused by locks