基于事务和对象语义的长事务处理方法

Long-living Transaction Processing based on Semantics of Transactions and Objects

数据库应用中出现的长事务严重降低了系统性能,一些研究者提出利用事务的语义信息对事务进行分解来解决这个问题,但是由于分解本身的需要满足一些特性,使得一些长事务无法分解。本文在事务分解的基础上提出了利用对象的语义信息来提高事务的并发度的方法,给出基于语义的正确性标准和一种基于有向图的正确性判定方法。

The system performance is decreased greatly by the appearance of long-living transaction in the application of databases. Some researchers proposed the decomposition of transactions to solve this problem by using the semantic information of transactions. But there are still some long-living transactions that cannot be decomposed because of the decomposition itself, which should satisfy desirable properties with respect to the original collection of transactions. In this paper, based on the decomposition of transactions, the semantic of object is exploited to improve the concurrency of transactions. The new semantic-based correctness criterion is given and a deciding method based on directed graph is proposed.