摘要：When multiple instances of a database application (DBA) are executing concurrently, transactions from these instances interleave. Due to unintentional grouping of operations of the DBA into transactions, some interleavings of these transactions may cause offline concurrency problems. Although these problems are analogous to lost update, dirty read, non-repeatable read, and phantom problems, the concurrency controller of a DBMS cannot ensure ACID properties because from the DBMS's point of view the transactions involved in the offline concurrency problems are executed successfully. This paper gives the enumeration of patterns of interleaving of operations which identify potential offline concurrency problems in a DBA.
